performer receipts. The discrepancy between obligations and performer receipts in fiscal year (FY) 2001 amounted to a hefty $7.3 billion—smaller than the discrepancy between authority and receipts, but still worthy of attention. In an ideal world, the reports of spending and receipts would match, perhaps with some adjustment for lags in the transfer of funds.
There have been several attempts to disentangle this apparent discrepancy over the past several years (Congressional Research Service, 2000). This is considered an important issue by many. The discrepancy is annoying as an accounting anomaly to analysts who seek to trace the funding streams and answer the question “Where did the money go?” (Koizumi, 2003). It is also of concern to members of Congress and others who have oversight and legislative responsibilities, because the discrepancy casts a shadow over the accuracy of the data that are used to gauge the overall health and vitality of the nation’s R&D enterprise, as well as answer the question “Are spending and receipt estimates accurately depicting how the money is flowing, and are the funds flowing in amounts and directions prescribed by legislation?” (U.S. General Accounting Office, 2001).
The discrepancy is of special concern to the panel and others who have studied it because its existence may signal a systemic quality problem in one or more of the NSF datasets. One litmus test of the quality of any statistical series is to compare the series with comparable data collected by other means. It is common statistical practice for federal statistical agencies to use administrative data sources to benchmark the totals derived from surveys. Measurement of undercoverage and overcoverage error (the failure to include or exclude some population units in the frame used for sample selection) often rely on comparing survey estimates to independent sources (U.S. Office of Management and Budget, 2001).
Prior to 1980, the discrepancy was quite small and generally indicated that federal obligations exceeded performer spending.
In the 1980s, the discrepancy became noticeable and changed, and for nearly a decade, spending appeared to significantly exceed obligations.
Around 1990, the trend shifted again, and spending once again has lagged obligations and outlays, only now by fairly substantial amounts.
The timing of these swings in the data series is in close alignment with changes in the design and operation of the NSF data collections, so it is