What the overhead ratio can and cannot tell us
Dan Pallotta asked us to look past overhead in his 2013 talk. Current returns let us test the reporting measure, while leaving his larger questions about pressure, ambition, and results open.
By the CharitySense Data team
Published Updated
01/question Pallotta left us
The question Pallotta left us
In his 2013 TED talk, Dan Pallotta argued that nonprofits face a different set of expectations about pay, marketing, risk, time, and access to capital. His challenge to the overhead reflex has lasted: why do we often ask what share of spending goes to administration before asking what the work achieves?
We can read expenses from a return. We cannot read donor tolerance for a larger budget, or the value of a program to the people it reaches. Our narrower test is whether the familiar program-expense percentage says as much as readers sometimes ask it to say.
A functional-expense share describes how costs were reported. The return does not turn that share into a measure of results.
02/First choose returns
First, choose returns that can support the ratio
We began with the latest available full Form 990 for each of 485,109 organizations. Of those, 277,653 have a latest full return for tax year 2024. For a program share, we require a positive total expense and an explicitly reported program-service amount between zero and that total. This leaves 259,132 comparable returns, each from a different organization.
Another 15,514 returns with positive expenses have no program amount. We do not turn a blank into zero. We also set aside 152 returns where the reported program amount lies outside zero to 100% of total expenses, plus returns with zero or negative total expenses. Those cases remain part of the filing record, but an ordinary percentage chart would disguise their different measurement problem.
Each bar groups organizations by their reported program-service expense divided by positive total expense. The endpoint bars are exact values. Select a bar for counts and a second expense measure.
| label | count | Percent | median Expense | event Expense |
|---|---|---|---|---|
| 0% | 16,494 | 6.4 | 246,827 | 1,369 |
| 0–20% | 3,849 | 1.5 | 195,648 | 617 |
| 20–40% | 5,512 | 2.1 | 232,362 | 1,033 |
| 40–60% | 13,842 | 5.3 | 355,152 | 3,362 |
| 60–80% | 58,802 | 22.7 | 765,426 | 16,779 |
| 80–<100% | 129,653 | 50 | 716,010 | 27,706 |
| 100% | 30,980 | 12 | 235,168 | 2,804 |
The broad middle matters. Exactly half of these returns fall between 80% and less than 100% program share. Another 22.7% fall between 60% and less than 80%. At the endpoints, 16,494 report exactly zero and 30,980 report exactly 100%, together 18.3% of the comparable group.
03/perfect percentage still has
A perfect percentage still has a second expense line
Among the 30,980 returns at exactly 100% program share, 2,804 also report a positive direct expense for fundraising events on their supplemental event schedule. That is 9.1% of the exact-100% group. The IRS instructions place direct event costs in the revenue statement rather than the functional fundraising-expense column. A perfect program share on one part of the form can therefore sit beside a real reported event cost elsewhere.
This overlap is a floor, because the supplemental schedule is subject to filing thresholds. It also does not tell us what an event raised, whether the expense was reasonable, or whether the program was effective. The sharper observation is about the boundaries of a ratio. Our separate look at zero fundraising expense follows this reporting distinction in a different group of 2024 returns. Its percentage and this one have different denominators and should not be compared as a trend.
04/larger argument reaches beyond
The larger argument reaches beyond the form
Pallotta also asked whether the sector can build organizations at the scale of its ambitions. We can measure concentration, but concentration alone cannot tell us which organizations wanted to grow or why some did not. In a separate analysis of 790,675 organizations with comparable latest full or short-form revenue, we found the largest 1% accounted for 66.8% of reported revenue. Those latest returns span tax years, and that group excludes postcard-only filers and private foundations. It describes the distribution of dollars, not the cause of that distribution.
Compensation is visible in another part of the return, but donor reaction is not. A line for fundraising expense cannot reveal the advertising that was never attempted. Assets tell us what an organization had at year-end, while an unfunded long-term plan leaves no comparable entry. Our compensation analysis explores one of those visible pieces with its own definitions.
Research outside the filings has tried to understand the pressure behind these choices. A 2009 study of four youth-serving nonprofits and interviews with funders described how expectations about low overhead can affect investment in organizational infrastructure. Its evidence is qualitative and specific to its sample. Our return counts do not independently establish that cycle across the sector.
05/What we would ask
What we would ask alongside the percentage
The ratio is still a useful invitation to look more closely. We would ask how the organization allocates shared work, whether an event has direct costs reported elsewhere, and what its program actually does. A hospital's expense mix will answer a different question from a small grantmaker's. For groups with filed relationships or federal awards, those records can add context about counterparties and public funding. Neither source supplies a universal outcome score, and neither covers every organization.
That is why the most interesting comparison may be among genuine peers with similar work and reporting obligations. An unusual percentage can then become a question for a return and its notes, rather than a shortcut to a verdict. Our diligence guide offers a way to read several kinds of evidence together.
Behind the Data
We read each organization's latest available full Form 990 in the current archive on September 29, 2026, then restricted the chart to organizations whose latest full return is for tax year 2024. The chart divides the explicitly reported program-service expense by positive reported total expenses. It excludes blank program values, zero or negative total expenses, and ratios outside zero to 100%. Each organization appears once. The seven chart bands do not overlap; exact zero and exact 100% are separate endpoints. Percentages use 259,132 comparable organizations as the denominator and are rounded to one decimal place.
For the event cross-check, we counted a positive direct fundraising-event expense on the supplemental schedule among the exact-100% group. A missing schedule cannot establish an absence of event costs. We also counted the reported states of the management and fundraising columns, without inferring why a filer used them as it did. The concentration figure comes from a separate latest-return analysis of full and short-form filers and has its own population. Recent filing years are still arriving, and churches generally are not required to file a Form 990. These are measurements of reported expenses, not an assessment of impact. See our methodology for the archive's broader scope.
Sources
- CharitySense Data, read-only analysis of 485,109 latest available full Form 990s, including 277,653 latest tax-year-2024 returns; checked September 29, 2026.
- Dan Pallotta, “The way we think about charity is dead wrong” (TED, 2013); TED's contemporary account of the five themes.
- Internal Revenue Service, 2024 Form 990 instructions, especially the functional-expense columns and direct event expense rules.
- Ann Goggins Gregory and Don Howard, “The nonprofit starvation cycle”, Stanford Social Innovation Review (2009).
Ask the research team
What would you want us to pull from the filings? Tell us, and a future piece may answer it.
Final Thoughts
The current returns show both why people reach for a program percentage and why they should keep reading after finding one. Most comparable returns report a share between 60% and 100%; the exact endpoints still need context, and a perfect functional split can coexist with event costs elsewhere on the form. Pallotta's broader question about what nonprofits could accomplish with different expectations deserves evidence beyond an expense allocation. We can begin with the filing, then ask what it leaves open.
About the author
The CharitySense Data team
Research team
We read public nonprofit filings and other sources to understand how the sector works. This blog shares what we find, along with the limits of the evidence. Each article explains how its figures were measured and cites the records or primary sources behind them.