Open data on public expenditure — datasets, categories and a SPARQL endpoint for research and analysis

Measuring US Federal Payment Integrity Through Public Records

Federal payment integrity is often discussed through a single headline figure: the estimated value of improper payments. That number matters, but it does not fully describe how public money moves, where controls fail, or whether agencies correct errors after they occur. A stronger assessment combines official estimates with transaction-level records, award data, agency reports, and repeatable calculations.

Publicspending.net provides a useful setting for this work because it brings public payment records into a searchable and comparable environment. Its datasets, standardized fields, graphs, bulk downloads, SPARQL endpoint, and ontology-based tools can help researchers examine federal expenditure patterns without depending on one agency’s presentation of its own results.

The aim is not to label every unusual payment as waste or fraud. Public records rarely contain enough context to establish intent. They can, however, reveal measurable signals: missing recipient information, repeated transactions, rapid changes in spending, unusually concentrated awards, payments made after a contract ends, and differences between reported obligations and disbursements.

What Payment Integrity Measures

Payment integrity covers the full path from an approved obligation to a completed disbursement. A payment may be improper because it went to the wrong recipient, exceeded an authorized amount, lacked required documentation, duplicated another payment, or supported a person who was ineligible for a program. Fraud is one possible cause, but administrative error, weak data matching, and unclear eligibility rules are also common explanations.

The most familiar official measure is the improper payment rate. Agencies generally estimate the share of payments that should not have been made or were made for an incorrect amount. The estimate can include overpayments, underpayments, payments to ineligible recipients, and payments lacking sufficient supporting evidence. Its value depends on the quality of the agency’s sampling design, definitions, and reporting period.

Public-record analysis adds a second layer. Instead of asking only how much an agency estimates it paid improperly, researchers can measure observable risk indicators across records. These indicators do not replace audits. They show where an audit, records request, or program review may produce the greatest value.

Useful measures include:

The Public Records Behind the Analysis

The Federal Awardee Performance and Integrity Information System and federal contract records can provide details about vendors, awards, modifications, competition, and performance history. USAspending.gov brings together federal award information under the DATA Act, including contracts, grants, loans, and other assistance. These sources are valuable for linking payments to the awards and recipients that authorized them.

Agency financial reports add information that transaction files may not show. Annual financial reports and improper payment reporting can explain program coverage, estimation methods, corrective actions, and known limitations. Treasury payment and spending data can support analysis of disbursement timing and aggregate cash movement, although the level of detail differs from award-level records.

Publicspending.net can help align these materials by standardizing names, dates, amounts, jurisdictions, and organizational relationships. That alignment matters because the same recipient may appear under several legal names, abbreviations, or identifier formats. A vendor-level payment total is only as reliable as the matching process behind it.

Researchers should preserve source provenance. Every calculated metric should identify the source file, reporting period, field definitions, filters, currency treatment, and exclusions. A result that cannot be reproduced is difficult to audit, even when the underlying records are public.

Metrics That Can Be Calculated Reliably

A payment integrity dashboard should begin with transparent descriptive measures. Total obligations, total outlays, transaction counts, average payment size, median payment size, and spending by agency or program establish the scale of activity. These figures help prevent a small number of striking transactions from distorting the overall picture.

Duplicate-payment screening is another practical method. A basic screen can group transactions by recipient, amount, date, agency, and award identifier. Exact matches deserve review, but near-duplicates can be more informative: two payments to the same recipient within a short period, for the same amount, under related award numbers. The result is a review queue, not proof of an improper payment.

Concentration measures show dependence on major recipients. The share of total spending received by the largest recipients can be calculated alongside the Herfindahl-Hirschman Index or a simple concentration ratio. High concentration may be appropriate for large infrastructure, defense, or health programs, so the measure must be interpreted with procurement type, contract size, and market structure in mind.

Timing metrics can reveal control weaknesses. Analysts can compare award dates, performance periods, modification dates, invoice dates, and payment dates. Payments after an award’s stated end date may reflect a legitimate closeout or modification, but they can also indicate stale data or a control exception. A useful dashboard reports both the number of cases and their dollar value.

Comparing Official Measures With Observable Signals

Official improper payment estimates and public-record indicators answer different questions. The official rate is generally designed to estimate program error under a defined methodology. A public-record signal describes what can be observed in available data. Treating the two as interchangeable creates false precision.

The comparison below illustrates how the measures can work together.

Measure What it captures Typical source Main limitation
Official improper payment rate Estimated share of payments that were improper Agency financial reports and required government reporting Methodology and coverage vary by program
Duplicate-payment indicator Transactions with matching or near-matching attributes Award and payment records Similar transactions may be legitimate
Recipient concentration Dependence on major vendors or beneficiaries USAspending and agency expenditure data Concentration can reflect lawful market structure
Post-period payment flag Payments recorded after an award or performance period Award dates and disbursement records Amendments and closeout activity may explain the timing
Unmatched payment share Records that cannot be linked to an award, recipient, or program Integrated public datasets Missing identifiers may result from technical limits
Recovery or recapture rate Amount identified and recovered after an error Agency reports and recovery data Recovery practices and reporting completeness differ

A high public-record risk score should therefore be described as a prioritization signal. It does not establish an improper payment rate, and it should never be presented as a fraud estimate without additional evidence. The most responsible analysis publishes the rule used to flag a transaction and shows how results change when that rule is tightened or relaxed.

Making Federal Data Comparable

Comparability is one of the hardest problems in government expenditure research. Agencies may define an obligation, payment, award, recipient, and program year differently. One dataset may report a transaction as a modification, while another aggregates it into the parent award. Fiscal years can also complicate comparisons because the federal fiscal year runs from October 1 through September 30.

A consistent data model should separate at least four concepts: the legal recipient, the award or assistance instrument, the financial event, and the program or agency responsible for the spending. It should also retain original identifiers rather than replacing them with a single standardized label. This makes it possible to trace a cleaned record back to its source.

Inflation and currency are less difficult for federal-only analysis than for international comparisons, but dollar values still need context. Nominal spending in different fiscal years should not be compared without an adjustment or an explicit statement that values are unadjusted. Large awards should also be examined separately from routine administrative payments because their scale can dominate aggregate results.

Publicspending.net’s ontology-based approach is useful when analysts need to connect concepts across datasets. A query can distinguish an agency from a bureau, a recipient from a parent company, or a payment from an award. That semantic structure supports cross-jurisdiction research while keeping the definitions visible.

Interpreting Anomalies Without Overclaiming

An anomaly is a record or pattern that differs from an expected baseline. It may be unusually large, unusually frequent, unusually timed, or inconsistent with neighboring records. The baseline should be specific. Comparing a defense procurement transaction with a small grant payment is unlikely to produce a meaningful integrity test.

A sound workflow compares like with like: the same agency, program, award type, recipient category, and time period where possible. Analysts should calculate rates as well as counts. Ten flagged transactions may sound serious, but their significance changes if they represent ten out of one hundred payments or ten out of ten million.

Human review remains essential. A flagged payment may be explained by an approved modification, a recurring benefit, a fiscal-year transition, a pass-through arrangement, or a data-entry convention. Researchers should record the explanation, preserve competing interpretations, and avoid publishing personal information that is not necessary for accountability.

The strongest public spending investigations combine several signals. A recipient with high spending concentration, repeated amendments, weak award linkage, and unusual payment timing merits closer examination than a recipient triggering only one broad rule. Even then, the proper language is “requires review” or “presents an observable risk pattern,” not “committed fraud.”

Building A Reproducible Integrity Dashboard

A useful dashboard should make its calculations inspectable. Users need to know the reporting period, included agencies, award categories, treatment of refunds and adjustments, duplicate-detection rules, and method used to handle missing values. Downloadable records and query examples allow independent researchers to test the findings.

Recommended dashboard components include:

Visual design should distinguish confirmed administrative findings from analytical flags. Color-coded risk categories can be useful, but they should not imply guilt. A separate view for data quality—missing identifiers, inconsistent dates, and unmatched awards—helps users understand whether a pattern reflects spending behavior or limitations in the source data.

The dashboard should also support historical revision. Federal records can be corrected, amended, or republished. Keeping version dates and change logs prevents researchers from treating a later correction as an unexplained contradiction.

Turning Open Data Into Better Oversight

Public records cannot reveal every control failure. Some improper payments are visible only in case files, eligibility systems, invoices, or internal audit samples. Still, open data can make oversight more systematic by showing where public resources are concentrated and where transaction records diverge from expected patterns.

Federal payment integrity metrics become most useful when they connect scale, quality, timing, and accountability. Official agency estimates provide program-level context. Standardized payment data shows how money moves. Award and recipient information makes relationships visible. Reproducible calculations allow journalists, auditors, researchers, and citizens to examine the same evidence.

Explore the federal spending datasets and query tools on Publicspending.net, compare agency and recipient patterns across reporting periods, and use the source records to build a documented review of how public money is paid, recorded, and monitored.