Scientific Misconduct: How NIH Built Its Estimate

This article explains how the U.S. National Institutes of Health (NIH) arrived at its widely cited estimate that scientific misconduct costs the U.S. government roughly $2.5 million per case and over $100 million annually in direct research funding. The estimate comes from a 2013 analysis that combined the prevalence of misconduct with the cost of a single funded research project. You will learn the data sources, the calculation steps, and the limitations of that figure.

Formula

The annual cost of scientific misconduct is calculated as:

Annual Cost = Cost per Case × Number of Cases per Year

Where:

• Cost per Case (USD) – the average direct cost of a single NIH-funded research project, taken as $2.5 million in the NIH estimate.
• Number of Cases per Year – the estimated annual number of NIH-funded misconduct cases, taken as 43 in the NIH estimate.

The result is the estimated direct cost to NIH of scientific misconduct in a given year.

Worked Example

Using the default values from the calculator (cost per case = $2,500,000, cases per year = 43):

1. Multiply the cost per case by the number of cases: $2,500,000 × 43 = $107,500,000
2. The estimated annual direct cost is $107.5 million.

This matches the NIH estimate of approximately $100 million per year.

Background: Why NIH Needed an Estimate

Scientific misconduct—fabrication, falsification, or plagiarism in research—undermines trust in science and wastes public money. The NIH, as the largest public funder of biomedical research in the U.S., needed a concrete number to justify oversight and to communicate the scale of the problem to policymakers. Before 2013, no official estimate existed; most discussions relied on anecdote or small surveys.

How NIH Built the Estimate

The estimate was published in a 2013 editorial in Nature by NIH officials, including then-Director Francis Collins. The calculation used two main inputs:

• Prevalence of misconduct: They cited a 2009 meta-analysis by Daniele Fanelli that found about 2% of scientists admitted to fabricating or falsifying data at least once. This was the best available figure at the time.
• Cost per case: They used the average cost of an NIH research grant, approximately $2.5 million over its lifetime, as a proxy for the cost of a single misconduct case.

They then multiplied the number of active NIH-funded researchers (about 200,000) by the 2% prevalence to get approximately 4,000 researchers who might commit misconduct. However, they acknowledged that only a fraction of those are detected, so they used a more conservative estimate of 43 cases per year—a number derived from the actual number of misconduct findings reported to NIH by institutions each year.

How to Read the Result

The $100 million figure represents direct costs only—the money NIH gave to the fraudulent projects. It does not include indirect costs like wasted time of other researchers, retraction costs, or damage to public trust. It is also a snapshot from 2013; the actual number could be higher or lower today. The estimate is an order-of-magnitude indicator, not a precise accounting.

Common Mistakes and Limitations

• Confusing direct and total costs: The $100 million is direct grant money, not the full economic cost, which could be several times higher.
• Assuming 2% of researchers are caught: The 2% is self-reported admission, not detection rate. The actual number of confirmed cases is much smaller.
• Ignoring that the prevalence estimate is contested: Some studies suggest lower or higher rates; the NIH estimate is sensitive to this input.
• Treating the figure as current: The estimate is from 2013; inflation and changes in research funding have not been updated.

Alternatives and Updates

Since 2013, no official NIH update has been published, but other researchers have attempted more comprehensive estimates. For example, a 2015 study in Proceedings of the National Academy of Sciences estimated that U.S. federal agencies spend about $1.5 billion per year on research that is later retracted due to misconduct, though that includes indirect costs. The U.S. Office of Research Integrity (ORI) also tracks confirmed cases, but those numbers are small (typically 10-20 per year) and do not account for undetected misconduct.

Frequently Asked Questions

How did NIH get the 2% prevalence figure?

The 2% comes from a meta-analysis by Daniele Fanelli, published in 2009, which pooled surveys of scientists. It found that about 2% of scientists admitted to fabricating or falsifying data at least once. This is a self-reported rate, so it likely underestimates the true prevalence.

Why did NIH use 43 cases per year instead of 4,000?

Because 2% of 200,000 researchers would be 4,000, but that number assumes every admitted misconduct is detected and confirmed. In reality, only a small fraction is caught. The 43 is the average number of misconduct findings that institutions report to NIH each year, so it is a conservative, evidence-based figure.

Does the $2.5 million per case include the cost of investigating?

No. The $2.5 million is the average cost of an NIH grant, not the cost of investigation. Investigation costs (legal, administrative) are additional and not included in the NIH estimate.

Is the $100 million figure still accurate today?

It is the most recent official estimate, but it is from 2013. Adjusting for inflation, it would be about $130 million in 2025. However, no updated NIH calculation has been published, so it remains the cited figure.

What does the NIH do with this estimate?

NIH uses the estimate to justify its oversight policies, such as requiring institutions to have research integrity programs. It also helps Congress understand the stakes when funding research.

Why doesn’t NIH just count confirmed cases?

Confirmed cases are rare (about 10-20 per year), but they represent only the tip of the iceberg. The estimate uses prevalence data to capture the hidden cost, which is more useful for policy planning.

Related

• Make Pinterest Board Private: Desktop & Mobile
• Get Personal Folders Back in OS X Lion Sidebar
• How Is Pace Calculated? Running Pace Explained

in t | Total direct research cost per case | USD | 2500000 | 0
in n | Number of misconduct cases per year | - | 43 | 0
out c | Annual cost of misconduct | USD | t * n | 0