Nonprofit statistics from public filings
We expected the size of the archive to be the surprise. The deeper surprise is how the picture changes when a filing, an organization and a public record are counted separately.
By the CharitySense Data team
Published Updated
01/One organization can leave
One organization can leave several kinds of record
The detailed analysis in this article uses a September 2026 snapshot, computed September 26 and checked September 29: 2,222,250 organization profiles and 8,868,546 filing records. A charity may file returns from multiple years or an unrelated-business return alongside its annual information return. The totals describe different units, so dividing the filing count by the profile count would tell us little about a typical organization.
The five forms also differ in what they ask. That snapshot includes 1,540,498 e-Postcard notices, or 17.4% of its filing records. The notice asks for eight basic items, not a program-expense statement. A comparison of program spending should leave those organizations out rather than treat the missing amount as zero. We walk through the full form mix in our essay on public 990 search.
The first useful question is often: what is the thing we counted?
02/most recent full return
The most recent full return has its own calendar
In the September 2026 analysis, we selected one latest full Form 990 or short Form 990-EZ per organization, giving us 803,975 organizations. The most common tax year in this defined group is 2024: 403,785 organizations, or 50.2%, have it as their latest available full or short return. Another 156,572 have a 2025 return. The newest tax years continue to arrive, so this is a dated snapshot rather than a live count.
Organizations, by latest available Form 990 or 990-EZ tax year. Select a bar to split the two forms.
| year | full | short | total |
|---|---|---|---|
| Through 2019 | 42,504 | 81,772 | 124,276 |
| 2020 | 7,911 | 14,234 | 22,145 |
| 2021 | 4,466 | 11,773 | 16,239 |
| 2022 | 8,555 | 18,451 | 27,006 |
| 2023 | 27,935 | 26,017 | 53,952 |
| 2024 | 275,073 | 128,712 | 403,785 |
| 2025 | 72,261 | 84,311 | 156,572 |
There is a small reversal inside the newest year. Among organizations whose latest full or short return is for 2025, 84,311 have a short return and 72,261 have a full one. For 2024, full returns outnumber short returns by more than two to one. We cannot read that shift as a change in the sector's makeup: the two tax years have had different amounts of time to enter the archive.
We also found 189,666 organizations, 23.6% of this cohort, whose latest full or short return is from before 2023. That does not prove they stopped filing. An organization can appear later on another form, and the public archive itself has limits. The year chart records the newest full or short return we found for each organization, not a verdict on its current activity.
03/average changes when
An average changes when the form changes
Revenue is available for 790,675 of the organizations in that comparison. Their median reported revenue is $122,732; the arithmetic mean is $3.94 million. A few very large institutions pull the mean far above the middle organization. The two numbers are both correct for this cohort, yet they answer different questions.
Form choice shifts the middle again. Among 438,443 organizations with revenue on a latest full return, the median is $420,678. Among 352,232 with a latest short return, it is $46,147. Short-return eligibility is tied to organizational size, so these are different filing populations. We follow the upper tail and its revenue share in the shape of nonprofit money.
04/richer record has uneven
The richer record has uneven edges
In the September 29, 2026 inventories, filings were accompanied by other public evidence, each with a different purpose: about 26,900 organization-level federal-award records, about 456,000 records describing an organization's side of known relationships, and a separate, selected set of about 33,500 website evidence records. These are record inventories, not totals of awards, relationships or people helped.
The mismatch is informative. A website capture is far rarer than a public profile, but its absence does not mean an organization has no website or no program. A relationship record can expose a named connection without proving that every transfer has been mapped. An award record reflects a government source and time frame rather than an IRS return. We examine how these distinctions affect real decisions in our grantee diligence essay.
05/map fades at
The map fades at the edges of filing rules
We can learn a great deal from millions of returns, but filing visibility varies. Churches and certain church-affiliated organizations generally do not have to submit annual returns or notices. A small organization may submit only the brief e-Postcard. A 2025 return can still enter the archive after the September 2026 snapshot. These are properties of the evidence, not shortcomings of the organizations.
The result is a richer, less tidy picture of the nonprofit world. The same archive can support a broad count of profiles, a narrow comparison of revenue, or a close look at a named relationship. Each view needs its own denominator and date. Our faith and geography analysis shows why a filing-based map should never be mistaken for a census of all religious activity.
06/Behind the Data
Behind the Data
We checked the public-site statistics on September 29, 2026. They were computed September 26 and report 2,222,250 profiles, 8,868,546 filing records and 1,540,498 e-Postcards. The five published form counts sum to the filing total; 1,540,498 divided by 8,868,546 is 17.4% after rounding. Filing records can repeat an organization across years and forms. The live profile and filing totals at the top of this page come from the separately timestamped published-site statistics.
For the September 2026 tax-year chart, we read every full and short return in that snapshot and selected one per organization: the higher tax year wins, with the full return winning a same-year tie. The resulting 803,975 organizations are the denominator for year shares. The “through 2019” bar combines all earlier tax years; 2020 through 2025 appear separately. The before-2023 figure combines 2007 through 2022. This selection ignores a later e-Postcard or private-foundation return, so the chart does not establish whether an organization is currently filing.
Of those selected organizations, 790,675 have numeric revenue. We calculated mean and median from those values, retaining filed zero and negative amounts. The full-return and short-return medians use their respective revenue-bearing subsets. Additional evidence counts are from the September 29 inventories, rounded only where the article says so. They are not complete-coverage claims. The website set is gathered separately and can change between reads.
Sources
- Our current published-site statistics and read-only analysis of latest full and short nonprofit returns, plus separate public-evidence inventories, checked September 29, 2026. See the methodology.
- IRS, Form 990-series downloads and annual return descriptions.
- IRS, e-Postcard guidance and annual filing exceptions.
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
We came away with a more useful picture than a single sector average. There are millions of public records, but each form makes a different slice visible. The financial middle changes with the filing group. Newer sources add texture while leaving gaps we need to name.
The next time a nonprofit statistic feels surprisingly simple, we will look for its unit and its denominator first.
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.