NSF Sampled 1.4% of Cornell's Grant Spending. It Found Problems.

The National Science Foundation's Office of Inspector General⧉ -- NSF's independent internal watchdog, which hired outside accounting firm Sikich CPA LLC to do the fieldwork -- audited how Cornell University spent $349.9 million⧉ it charged to 781 NSF research awards between October 2021 and September 2024.▸ Auditors didn't check all of it, or even most of it: they judgmentally selected 92 transactions worth $4,784,824 -- about 1.4% of the total -- and asked Cornell to document each one. Inside that small slice, they questioned $121,981⧉ across 23 awards. Cornell disputes more than half of it -- and the other 98.6% of what it charged NSF over three years was never examined by this audit at all.

Cornell is a major recipient of NSF money -- in fiscal year 2024 alone, the university reported $1.2 billion⧉ in grant and contract revenue, 74% of it from federal, state, and local government.▸ NSF OIG's audit of $349.9 million of that money charged to NSF awards specifically found 30% went to salaries and wages, another 24.5% to Cornell's own indirect ('overhead') costs, and 14.5% passed through to subaward partners. To check whether that spending followed federal rules, auditors didn't examine the whole population -- they picked 92 transactions worth $4.78 million by judgment, not random sample, and asked Cornell to produce supporting documentation for each one. That is the entire scope of what this audit can tell anyone about Cornell's NSF spending: what happened in roughly 1 of every 73 dollars.

Even at that scale, the sample wasn't clean.▸ Auditors questioned $121,981 -- about 2.5% of the dollars they tested -- spread across four kinds of violations of 2 CFR Part 200⧉, the federal cost-principles regulation that governs whether a university can legally bill a research grant for something: $64,644 in outright unallowable costs, $32,128 the university couldn't adequately document, $22,321 billed to the wrong award, and $2,888 charged at the wrong overhead rate. The specifics read like a line-by-line grant ledger: a $479 business meal double-counted against a per diem, $576 in alcohol and a $408 cancellation fee on one trip, a $3,516 airfare seat upgrade and rebooking charged to federal money for personal reasons, and an H-1B long-term visa cost that auditors say should never have gone on an NSF award at all.

The smallest finding is the clearest illustration of how a rate mismatch turns into real money.▸ NSF's I-Corps⧉ program -- a $50,000 innovation-training award with an 11.11% overhead cap baked into its own solicitation -- had Cornell's full negotiated indirect-cost rate, 57%, applied to it instead. That single mismatch on one transaction cost $2,295 in questioned indirect charges. On a $50,000 award, the difference between 11.11% and 57% overhead is the difference between $5,000 and $28,500 -- money that, applied wrong even once, comes straight out of the direct research budget an I-Corps team was supposed to spend on its actual project.

Cornell did not simply accept the findings.▸ The university agreed to repay $54,438 of the $121,981 -- less than half -- and pushed back point by point on the rest. On a $26,404 technician-services charge auditors called inadequately documented, Cornell argued the vendor's invoice matched its original quote. On $17,330 in publication costs auditors say should have been split across co-funded projects, Cornell argued the other funding sources didn't benefit enough to warrant sharing the bill. Auditors' written response, included in the same report, didn't move: 'our position regarding this finding has not changed,' on more than one dispute. NSF's Director of the Division of Institution and Award Support now has a 6-month resolution clock under OMB Circular A-50⧉ to work through all 17 recommendations and decide which of the disputed $67,543 Cornell actually has to pay back.