What Nonprofits Actually Spend Money On

Donors ask what percentage goes to the cause. Boards ask whether administration is too high. Both questions assume there is a published number, and there is, but almost nobody who repeats it has looked at where it comes from.

So we went to the raw filings: 256,539 charities, every line of their expenses. The median organization spends 9.2% of its money on the back office, meaning rent, technology, office costs, legal, accounting, insurance and management fees combined. One in ten spends more than 31%.

The number you cannot get from the tax returns

Start with the thing that is usually skipped. The famous overhead ratio, the percentage split between programs, management and fundraising, is a functional classification. Form 990 Part IX asks organizations to take every expense line and allocate it across three columns: program services, management and general, and fundraising.

The free IRS data extract carries column A only, the total. The functional columns are not in it. So the percentage-to-programs figure cannot be computed from this source at all, and anyone who says they derived it from the IRS extract has derived something else.

That is not just a data limitation, it is a clue about the number itself. A functional allocation is a judgement. When a program director spends a morning on a grant report, the organization decides whether that morning is program or management, and two identical charities can book it differently and both be correct. The ratio is real accounting, but it measures a decision as much as it measures spending.

What the extract does carry is the natural breakdown: the same total, split by what the money actually bought. Salaries, rent, insurance, travel, depreciation. A natural category is a fact about a payment rather than a view about its purpose, which makes it harder to present favourably and more useful to compare.

Where the money goes, for the typical organization

These are median shares, so each column is the middle organization for that line, not a breakdown of one budget. The rows deliberately do not add to 100%.

The first row is the finding. The median charity spending between $25,000 and $250,000 a year reports nothing for salaries, nothing for rent and nothing for depreciation. Not a small amount. Zero.

59% of small charities have no payroll at all

That zero is not a filing gap. We checked: in every size band, the individual expense lines add up to the stated total for 100% of organizations, so the detail is being completed. These organizations genuinely paid nobody.

People cost includes salaries, officer compensation, pension contributions, other benefits and payroll tax. It rises to a peak of 53.0% between $5m and $25m and then falls above $25m, because the largest organizations spend a growing share on grants to other organizations and on contracted services rather than on their own staff.

The practical consequence is that comparing your staff share against a sector average is close to meaningless unless the average is size matched. An all volunteer organization and a $10m service provider are both normal, and they sit at 0% and 53%.

That total says nothing about how it is distributed, which is a separate question with a separate answer: 79% of organizations under $250,000 pay their officers nothing, and leadership pay runs from a median of $41,445 at the smallest organizations to $1,716,373 at the largest.

Administrative costs as a percentage, honestly measured

Here is the closest defensible answer to the administrative costs question, built from the natural lines a board actually argues about: occupancy, office expenses, information technology, legal fees, accounting fees, insurance and outside management fees.

Two things worth noticing. The median barely moves with size, from 10.0% to 7.4%, so there is no dramatic efficiency gain from scale in these categories. What collapses is the spread. The 90th percentile falls from 44.0% to 17.1%.

That is what size actually buys: not a lower cost base, but a narrower range of outcomes. A small organization renting an office in an expensive city, or paying for an audit it does not legally need, can find a third of its budget in this bucket with nothing wrong at all. Fixed costs do not scale down, and an organization spending $60,000 a year cannot spread a $12,000 audit across anything.

If you are using a threshold, note where the common rules of thumb land. The frequently quoted 15% and 25% administrative ceilings sit between the 50th and 75th percentile for small organizations and above the 75th percentile for large ones, so the same rule is lenient for a hospital and punishing for a food bank.

Almost nobody uses a professional fundraiser

The other half of the overhead argument is fundraising cost, and one part of it is measurable here. Form 990 has a separate line for fees paid to professional fundraising services.

Across the whole file, 9,182 organizations, or 3.6%, reported paying a professional fundraiser anything at all. The practice is rare and concentrated at the top, where one organization in nine uses one.

Read the fee column carefully, because it is a floor and not a commission rate. It is the fee divided by all contributions the organization received, including everything the fundraiser had nothing to do with. The consultant running a capital campaign is not taking 1.9% of the annual fund. The column tells you how much of total giving the line consumes, which is the question a board is actually asking, and the answer at every size is: much less than the folklore suggests.

Why the aggregate and the median disagree

One more caution, because it is the most common way these figures get misquoted. Add up every dollar spent by organizations under $250,000 and people cost is 19.3% of it. The median organization in that same group is at 0.0%.

Both are correct. A minority of small organizations employ someone, and those organizations spend more in total than the volunteer run majority, so they dominate the dollars while being outnumbered. Every sector wide expense average you see is one of these two numbers, and which one it is changes the answer completely. Use the median when you want to know what is normal for an organization, and the aggregate when you want to know where the sector’s money goes.

What to do with your own numbers

Take your statement of functional expenses, or Form 990 Part IX if you have filed one, and calculate three things.

People as a share of total expenses. Compare it only to your own size band above. If you are at 65% with a $2m budget you are at the top of the range but inside it, and that is a service delivery organization, not a problem.

The back office lines as a share of total. Occupancy, office, technology, legal, accounting, insurance, management fees. If you are past the 75th percentile for your size, the useful question is which single line is doing it, because in practice it is almost always rent or an audit rather than general sprawl.

The direction of travel over three years. A single year tells you very little. Costs that rise while the budget is flat are the signal, and no percentile table can see that for you.

Then read the result next to the two numbers that actually predict trouble, which are not on this page. The first is whether you spent more than you received, and 41% of organizations did. The second is how long your cash would last, where the median is five months and a third of the sector is under three. An organization with high administrative costs and a year of cash is fine. An organization with low administrative costs and three weeks of cash is not.

Method and limits

Source is the IRS Statistics of Income annual extract of tax-exempt organization financial data, Form 990, processing year 2024. It is a free public file covering every return the IRS processed that year, so it is a census and not a survey. The analysis script is published and the figures above can be reproduced from it.

The population is 501(c)(3) organizations with total functional expenses of $25,000 or more, which leaves 256,539 organizations. The floor removes dormant entities whose percentages are dominated by rounding.

Four limits are worth stating. Organizations filing Form 990-EZ or 990-N are not in this extract, so the smallest charities are underrepresented and the true share with no payroll across the whole sector is higher than 59%. Processing year is not tax year, so a minority of returns cover earlier periods. Functional allocation is absent, as set out at the top. And the natural categories reflect how each organization chose to code its own lines, which is more consistent than functional allocation but not perfectly consistent.

The single largest category on this page is the one we have not discussed: for the median organization under $250,000, 20.9% of spending lands in the Form 990 catch-all “other expenses” lines. That is where an organization with no staff and no office puts direct program costs, and it is a reminder that the tax return is a tax return, not a management account.