Testing a famous argument against the filings
What the 990s Say About the Overhead Question
In 2013, Dan Pallotta argued that the obsession with overhead is the wrong question. We processed millions of filings to see how much of that argument the data can actually carry.
We are checking his claims against the public record, not crowning him right or wrong.
of filings report a program share pinned at exactly 0% or 100%
A 2013 talk asked whether we judge charities by the wrong yardstick.
In his 2013 TED talk, “The way we think about charity is dead wrong,” Dan Pallotta argued that society holds nonprofits to a separate rulebook in five areas: compensation, advertising and marketing, the freedom to take risks, the freedom to take a long view, and access to profit and risk capital. He cited a few figures that stuck with us. Giving in the United States has hovered near 2% of GDP since the 1970s. From 1970 to 2009, by his count, only 144 nonprofits crossed $50 million in annual revenue while 46,136 for-profit companies did. And he pointed to large pay gaps between business-school graduates and charity executives.
His central complaint was about overhead. The reflex to ask “what percentage goes to the cause versus administration?” he argued, is the wrong question, and it starves organizations of the investment they would need to grow.
We process IRS Form 990 filings for a living, so we wanted to see how much of this argument the data can carry. We cannot test a TED talk. We can test specific, measurable claims against what the filings show, and we can be honest about the parts the forms simply do not cover.
A 990 records what an organization reported. It does not record what a donor would have tolerated, or what an organization would have done with capital it never had.
Which of his claims a tax form can actually speak to.
Before reading anything into the numbers, it helps to separate the claims we can check from the ones we cannot. Some of Pallotta's argument lives directly in the functional-expense section and the balance sheet. Some of it lives in attitudes, norms, and counterfactuals that no filing captures.
| Pallotta's claim | In the filings? | What we can check | What the data suggests |
|---|---|---|---|
| The overhead percentage is the wrong question. | We can read the program, admin, and fundraising splits straight off the functional-expense section of every 990. | More than a quarter of filings that report program spending sit pinned at a perfectly round 0% or 100%. The number is noisy enough that reading it straight off the form is unreliable. | |
| The sector cannot build to scale the way for-profits do. | We can measure revenue concentration and watch how the size distribution moved over a decade. | A few giants hold most of the dollars, and the typical organization stayed flat for the decade we observed. The data is consistent with a sector that struggles to grow the middle. | |
| Charities are held to a harmful rule against paying competitive compensation. | We can read reported officer and key-employee compensation, but not the social norms or donor reactions behind it. | The dollars are in the filing. Whether a pay level is punished by donors or boards is a question the 990 does not answer. | |
| Charities are discouraged from spending on advertising and marketing. | Fundraising expense appears on the form. Attitudes toward marketing do not. | We can see what was spent on fundraising, but the filing carries no signal about what an organization wanted to spend or was talked out of spending. | |
| Charities are denied a long time horizon and access to risk capital. | We can see assets, reserves, and growth over time, but not the counterfactual of capital that was never raised. | Balance-sheet strength is visible. The investments that did not happen because the capital was unavailable are, by definition, not in the record. |
Two of his five themes, the overhead question and the difficulty of building scale, sit squarely in the data. The other three involve what donors feel, what boards permit, or what investment never happened. Those we can describe around the edges, but we cannot settle them from 990s alone, and it would be dishonest to pretend otherwise.
The program ratio piles up at the edges, where it means the least.
The strongest thing our data has to say about Pallotta's thesis concerns the overhead percentage itself. If the program ratio were a clean measure of how well an organization channels money to its mission, we would expect it to spread across a sensible range. It does not. Among filings that report program spending, the distribution piles up at the extremes.
| Reported program share | Share of filings | What it usually means |
|---|---|---|
| Exactly 0% to program | About 9% of filings | Usually a blank or unreported program line that reads as zero, not an organization that spent nothing on its mission. |
| Exactly 100% to program | About 18% of filings | Usually every dollar placed in one box, or rounding, rather than an organization with literally no administrative cost. |
| Pinned at a round extreme | More than a quarter combined | A self-reported artifact pile-up at the very edges, which is where a single overhead ratio is least informative. |
About 9% of these filings report exactly 0% going to program, and about 18% report exactly 100%. More than a quarter sit pinned at a perfectly round extreme. An organization that genuinely spends nothing on its mission, or nothing on administration, is rare. A blank that reads as zero, a rounding habit, or every dollar dropped into a single box is common. These appear to be artifacts of self-reported data rather than descriptions of how the money actually moved.
When more than a quarter of the answers sit at 0% or 100%, a raw overhead percentage read straight off the form is a noisy signal, not a verdict.
This lends weight to Pallotta's “wrong question” point, though from a different direction than he took. He argued the overhead ratio is the wrong thing to optimize. The filings suggest it is also, very often, a number that cannot bear the weight people put on it. That is part of why we do not read a single efficiency percentage off the form when we compute ratings. The shape of the distribution is a warning label.
A few organizations hold most of the dollars.
Pallotta's second testable claim is that the sector cannot build the kind of scale for-profits can. We cannot reproduce his 1970-to-2009 count of organizations crossing $50 million, because our corpus is the modern electronic filing record rather than a five-decade longitudinal panel. What we can show is how concentrated the sector looks today.
| Segment | Share of organizations | Share of sector dollars |
|---|---|---|
| Top 1% by revenue | 1% of organizations | About 67% of sector revenue |
| Top 10% by revenue | 10% of organizations | About 93% of sector revenue |
| Organizations raising over $1B | About 0.1% of organizations | About 47% of the dollars |
| Organizations under $1M revenue | About four-fifths of organizations | About 2% of the dollars |
The top 1% of organizations by revenue hold about 67% of all sector revenue. The top 10% hold about 93%. Organizations raising more than $1 billion are roughly 0.1% of organizations but command about 47% of the dollars, a slice dominated by a handful of large hospitals and universities. Meanwhile the four-fifths of organizations under $1 million in annual revenue together hold about 2% of the dollars. The gap between the typical organization and the average one is just as stark: median organization revenue is about $208,000, while the mean is about $6.25 million, roughly thirty times larger, pulled up by those few giants.
This does not prove Pallotta's causal story. A concentrated sector could reflect many things besides a culture that punishes growth. But the picture is at least consistent with his claim that the middle of the sector struggles to build size.
The decade view
The totals roughly doubled. The typical organization did not.
Indexing to 2010 = 100 and following the years to 2020, total sector assets in our sample roughly doubled, from about $122 billion to about $306 billion, and total revenue climbed on a similar path. Over the same stretch, the median organization stayed flat and drifted down to about 81. The growth appears to have accrued to organizations that were already large.
One honest wrinkle in the decade trend.
We want to be careful here, because there is a measurement effect tangled up in the median. Over this decade, mandatory electronic filing pulled a long tail of small organizations into the visible record for the first time. When a flood of small filers enters the data, the median can drift down even if no individual organization shrank. So part of the flat median is real, and part of it is the record getting more complete.
That caveat cuts in an interesting direction. Even after accounting for the newly visible small filers, the totals still concentrated upward and the typical organization still did not grow. The trend Pallotta described is visible in the data. We just cannot attribute all of it to the cultural cause he proposed.
The typical organization stayed flat while the totals doubled. The data shows the pattern. It does not, on its own, explain why.
Where the filings go quiet.
Three of Pallotta's five themes run past the edge of what a 990 can tell us, and it is worth stating plainly where the data goes quiet.
On compensation, the filings do report officer and key-employee pay, so the dollar figures exist. What they cannot tell us is whether a given pay level was punished by donors, capped by a board, or quietly suppressed by the fear of looking lavish in next year's filing. The number is in the record. The social pressure around it is not.
On advertising and marketing, we can see fundraising expense, but the form carries no signal about an organization's appetite for marketing or about investment it talked itself out of. A low fundraising line could mean discipline, scarcity, or caution. The filing does not distinguish among them.
On time horizon and risk capital, we can read assets, reserves, and growth, but the investments that never happened because the capital was never available are, by their nature, absent from the record. Pallotta's argument is partly about a counterfactual, and counterfactuals do not file tax returns.
There is one more corner the filings barely reach. Religious organizations are only about 5% of the filings we see, because churches are not required to file a Form 990 at all. A large share of US giving sits largely outside the public filing record, so any read on the sector's overall behavior is a read on the part that files.
A better question than the ratio.
If the overhead percentage is a noisy number, the constructive move is not to throw it away but to ask what it was standing in for. People reach for the program ratio because they want to know whether an organization is healthy, growing, and channeling money toward its work. Those are answerable questions, and they live in the same filings, just not in a single percentage.
The functional-expense split is most useful in context. A high share of spending landing in administration and salaries can make sense for a hospital or a university, and looks very different for a small grassroots program. A program ratio pinned at 100% is more likely a reporting habit than a triumph. Read alongside revenue trend, reserve strength, governance fields, and how an organization compares with genuine peers, the same data starts to answer the question people were really asking when they reached for overhead.
The filings can tell you more than a single ratio ever could. The overhead percentage was never the best thing the form had to offer.
That is roughly where our reading lands. Pallotta's frustration with the overhead question is supported by the shape of the data. His claim about scale is at least consistent with how concentrated the sector looks and how flat the typical organization stayed. And his claims about compensation, marketing, and risk capital sit in territory the forms were never built to map.
A famous argument, measured against the public record.
Some of Pallotta's case shows up clearly in the filings. The overhead ratio piles up at the edges where it means the least, and the sector's dollars sit with a handful of giants while the typical organization stayed flat for a decade. Other parts of his argument live in attitudes and counterfactuals that a tax form does not record, and we have tried to say so rather than stretch the data to fit.
Hopefully, by reading the filings more carefully than a single percentage allows, more of the right questions can reach the organizations doing the work.
Final Thoughts
Testing an argument against data is not the same as judging the argument. The filings can confirm a pattern, narrow a claim, or mark the boundary of what is knowable, but they cannot reach a person's intent or a donor's reaction. We found that the overhead reflex is, by the shape of the numbers, a weak measure, and that the difficulty of building scale is at least visible in the record. We also found three claims the forms cannot adjudicate, and a large religious slice of giving that the forms do not see at all.
That mix feels like the honest answer. The most useful thing the data does here is not to settle the debate but to show which parts of it can be settled with evidence and which parts still rest on judgment.
Sources
- Dan Pallotta, "The way we think about charity is dead wrong," TED 2013
- IRS Form 990 resources and tools
- IRS Form 990 series downloads
- IRS Form 990-EZ, Short Form Return
- IRS exemption for churches and religious organizations
- Stanford Social Innovation Review, The Nonprofit Starvation Cycle
- CharitySense Data
