Two reporting chains for the same contract dollar, and the gap between them

The headline

These are not two measurements of one quantity. They are two submissions with different purposes, so the ratio measures reporting linkage — never error, and never missing money.

Whether linkage is falling cannot be read from these files

This page used to draw a trend line through the yearly figures. It no longer does, because the figures are not comparable with each other: File C states a fiscal year as of a submission period, the warehouse holds four snapshots of each year, and the snapshot published differs by year. A line through them measures which copy was opened.

Read this as a completeness indicator for the account-linkage submission, not as an audit finding about the underlying obligations. A dollar absent from File C is not a dollar that was not obligated; it is a dollar whose account linkage did not reach this dataset in the snapshot that was read. Controls FILEC-01 and FILEC-02 keep the whole series behind every figure on this page, and the linkage page shows the full grid of year against snapshot.

The full reconciliation

Why the two differ by construction: the award files record every FPDS contract action for the agency, while File C records the obligations that components mapped to a Treasury Account Symbol in their submission. Differences in timing, in modification handling, and in which components submitted complete account attribution all move the ratio. What the trend cannot be explained away as is a change in the underlying contract volume — action counts held near 4.4 million in every closed year.

Contract volume is steady; the File C row count is not the story either

A natural next thought is that linkage moves because there is more or less contracting to link. There is not: contract actions per closed year barely move. Nor does the File C row count explain the swing — the snapshots hold a comparable number of rows and wildly different dollar totals, which is why the dollar figure is the volatile one.

Closed fiscal years still move

Two warehouse vintages of the same award files, one month apart. If a snapshot's figures were stable, every bar here would be zero.

FY2021 closed five years ago and still gained $14.4M between these two extracts. FY2023 lost dollars and actions, meaning records were removed, not just revised. This is why every figure on this site is stamped with the vintage it came from: a number without one cannot be reproduced, and a disagreement between two numbers is a vintage difference until proven otherwise.

What would close the gap

Stated as work, not as a finding — this is the analysis a comptroller-side team would run next.

1. Decompose by component.

   File C attribution is submitted by component. Splitting linkage by awarding sub-agency identifies whether the fall is department-wide or concentrated in a small number of submitters, which changes the remediation entirely.

2. Separate timing from completeness.

   Half of this is now done: reading each fiscal year at every File C snapshot held shows the ratio moving by up to a factor of eight within one year, so within-year timing dominates. What remains is the other axis — re-running the same period across successive warehouse vintages, which would separate submission lag from a completeness failure. Neither can be read from a single published figure.

3. Test modification handling.

   Award files count every modification as an action. If File C attributes only base awards for some submitters, the two populations differ structurally and the ratio needs a like-for-like denominator.

4. Tie to the audit trail.

   Account linkage is what lets a contract obligation be traced to a Treasury account, which is the same evidence chain that supports the universe of transactions in the financial statement audit.