Charity overhead in official registers: what the filings show

Key findings

What share of their expenses do US charities report as program spending?

The median US charity that files a full Form 990 reports 86.1% of its expenses as program service expenses. The ratio is U-shaped by size: highest in the smallest band (90.0%), lowest for charities with USD 1 million to 10 million in revenue (84.2%), and back up to 88.1% above USD 100 million. Small charities spread much wider: in the smallest band the 10th percentile is 15.9%, in the largest it is 75.8%.

Program expense ratio = program service expenses / total expenses. US full Form 990 filers, FY2023, one filing per charity; data snapshot 21 October 2026. P10 to P90 are percentiles: 10% of charities in the band report a ratio below P10. "90%+" and "95%+" mean at or above that ratio.

Management and fundraising expenses by size

The median filer reports 10.9% of expenses as management and general and 0.0% as fundraising. More than half of all filers, 58.8%, report no fundraising expenses at all, and so do 45.7% of the 34,718 charities with total expenses above USD 5 million.

Filers from the table above whose return holds a reconciled Part IX split (columns B, C and D add up to column A). Shares are of total expenses.

Zero fundraising expense does not mean no fundraising. A charity funded by fees, government contracts or grants may have no fundraising cost, volunteer-run fundraising has no booked cost, and some costs of joint appeals can be allocated to programs. This data does not show which of these applies.

Which charity registers report a program, administration and fundraising split?

Two registers in GiveRadar's database hold the full functional split at scale. The US IRS Form 990 has all three figures for 250,956 of 257,755 full filers (FY2023), and Canada's CRA T3010 for 10,626 of 59,451 charities. Most other registers publish total income and expenditure but no split by function.

Most recent complete fiscal year per register: the latest year whose filing count reaches at least 95% of that register's highest yearly count. One filing per charity per year.

How to read the zeros. A zero means GiveRadar does not hold that figure. It does not mean the register never publishes it. England and Wales and Scotland follow the Charities SORP, which reports "charitable activities" and "raising funds" but has no separate administration line (support costs are allocated across activities), so an admin zero there is structural. The 990-EZ has no functional expense split (the 2 filings with one are tagging anomalies), and the 990-PF reports expenses in a different structure.

What do Canadian charities report on their T3010?

The median Canadian charity spends 85.0% of its total expenditure on charitable programs (CRA T3010, FY2023, 42,152 charities). The CRA's own fundraising ratio cannot be computed from this data: GiveRadar holds lines 5000, 5010, 5020 and 5100, but not the revenue lines 4500 and 4630 it needs.

Share of total expenditure (line 5100) spent on charitable programs (line 5000), FY2023.

Canadian program ratio by size

For the 10,537 charities that report program, administration and fundraising spending, the median program ratio is 78.6%, rising from 60.2% in the smallest band to 89.6% above CAD 100 million.

CRA filers with lines 5000, 5010 and 5020 all held, FY2023, CAD bands with the same thresholds as the US table. Ratio = line 5000 / line 5100.

Not comparable with the US table. Line 5100 includes gifts to other charities (line 5050), so grantmaking charities show a low ratio, which explains the low 10th percentiles. The CRA import also stores a reported zero as blank, so this group only contains charities with non-zero program, administration and fundraising spending: charities that do no fundraising drop out entirely, and zero-fundraising and zero-administration shares cannot be measured.

How we measured this

Every figure comes from register filings as held in the GiveRadar database on 21 October 2026. We counted organisations only; no individual charity is named or ranked here.

• Sources. IRS Form 990 e-file data (US), CRA T3010 open data (Canada), and the register filings listed in the coverage table.
• Fiscal year. Labelled by the calendar year in which it ends, as in the research note Charity finances 2021 to 2024.
• One filing per charity. Where a charity has several filings for a fiscal year, the one with the highest total revenue is used. Organisations GiveRadar has excluded as not being charities are left out.
• US year. FY2023. FY2024 holds 236,857 full Form 990 filers against 257,755 for FY2023 (91.9%), so FY2024 is not yet complete.
• US figures. Program, management and general, and fundraising expenses come from Part IX line 25 columns B, C and D, used only when B + C + D equals column A (total expenses) to the dollar. Where that check fails, the program figure falls back to the Part III line 4e total, kept only when it lies between zero and total expenses, and no management or fundraising figure is held. Total expenses = Part I line 18, which equals Part IX line 25 column A. Each figure comes from the same return; where an amended return exists, the later one is used.
• Revenue bands. Each charity is placed in a band by the filing's own total revenue for that year. Filers with revenue of zero or below fall in the lowest band.
• US exclusions. Of 257,755 full Form 990 filers for FY2023, 2,245 with total expenses of zero or below, 2,662 with no program figure held and 4 data quality outliers with a ratio above 100% (none below 0%) are left out. 252,844 remain. Outliers by band: under USD 100K 1, USD 100K to 1M 2, USD 1M to 10M 1, larger bands 0.
• Canada exclusions. For the program ratio by size, 54 charities with total expenses of zero or below and 35 outliers above 100% are left out, out of 10,626.
• Percentiles. Linear interpolation (PostgreSQL percentile_cont).

What this does not mean

A high program ratio is not proof of a good charity, and a low one is not proof of a bad one. For a donor-facing explanation, see what overhead percentages mean for donors.

• A program expense ratio is not a measure of effectiveness. It says how costs were labelled, not what they achieved.
• Reporting choices move the ratio. Charities decide how to allocate shared costs (staff, rent, IT) across functions, and many small filers book everything as program spending: 25.2% of the smallest US filers report exactly 100%.
• In-kind valuation affects it. Donated goods and services booked at market value raise program spending without any change in cash.
• Definitions differ by register. The US, Canadian and UK figures measure different things and should not be compared as if they were one scale.
• Small charities are under-represented. US filers under USD 200,000 in revenue can file the 990-EZ or the 990-N instead of the full 990, so the lowest band here contains only the small charities that chose or were required to file the full form.
• Late filings still arrive. FY2023 is complete enough to use, but late filings still come in.

Download the data

The aggregate tables behind this note are free to reuse under CC-BY-4.0. They contain counts, percentiles and shares per register, size band and group, with no per-charity rows.