NTEE Codes Explained: How US Nonprofits Are Classified

Almost every statement you have read about American philanthropy by sector — how much goes to health, to education, to the arts — rests on a classification system most people in the sector have never examined. It is called the National Taxonomy of Exempt Entities, or NTEE, and it assigns every tax-exempt organization a letter-and-number code describing what it does.

NTEE is genuinely useful. It is also, in a large minority of cases, wrong — because codes are frequently assigned once at formation, from a limited description, and then never revisited while the organization changes over decades.

If you are a foundation classifying grantees, a researcher aggregating giving by cause, or a nonprofit wondering why funders never find you, it is worth understanding both how the system works and where it breaks.

What NTEE is and who maintains it

NTEE is a descriptive taxonomy for classifying US tax-exempt organizations, developed to let regulators and researchers aggregate data for statistical analysis. It is used by the IRS and by the National Center for Charitable Statistics at the Urban Institute, which maintains the reference documentation.

An organization's NTEE code appears in the IRS Business Master File, and it is the field almost every "giving by sector" analysis depends on — including the cause pages on our own dataset.

Importantly, NTEE is descriptive, not regulatory. Your NTEE code does not determine your tax treatment, what you may do, or how you are supervised. It is a label for statistical purposes. That is precisely why it receives less attention than it deserves: nothing breaks immediately when it is wrong.

The structure

The code is hierarchical, working from broad to specific.

Major group — one letter, A through Z. Twenty-six groups covering the span of exempt activity: A for Arts, Culture and Humanities; B for Education; E for Health Care; P for Human Services; T for Philanthropy, Voluntarism and Grantmaking Foundations, and so on. NCCS in turn rolls these 26 letters up into ten broad categories — Arts, Education, Environment and Animals, Health, Human Services, International, Public and Societal Benefit, Religion, Mutual Benefit, and Unknown — which is why you will see both "26 groups" and "10 categories" quoted, correctly, for the same system.

Then digits, narrowing the activity. A two-digit number identifies a division within the major group, and further digits reach a specific activity. So B is Education, B2 is elementary and secondary education, and a fuller code such as B24 narrows further.

There is also a set of common codes that cut across major groups, covering functions rather than fields — research institutes, professional societies, fundraising entities, single-organization support groups. These attach the same meaning wherever they appear.

A revised format, NTEE-V2, restructures the same information for analytics, separating organization type from activity and exposing the higher-level industry code directly. If you are working with the data programmatically it is worth knowing both formats exist.

Where it goes wrong

Four failure modes, and they compound.

Assignment at formation, never revisited. A code is typically assigned when an organization applies for exemption, based on a brief description of intended activity. Organizations then run for decades and change substantially. A group that incorporated in 1994 to run after-school sports and now provides family housing support may still be coded under recreation.

Multi-purpose organizations get one code. An organization running a food bank, a health clinic and a job training program has one NTEE code. Two-thirds of what it does is invisible to any analysis using that field, and which third is captured is essentially arbitrary.

Grantmaking versus doing. Major group T covers philanthropy and grantmaking. An operating charity that also regrants may be coded T, or may be coded by its program area, depending on how it was described at formation. Analyses of "how much goes to grantmaking intermediaries" are unusually sensitive to this.

Codes are hard to change and there is little incentive. Updating a code requires a request to the IRS, and nothing bad happens if you do not — so nobody does.

The result is systematic rather than random error. NTEE tends to describe an organization's origin more accurately than its current work, which means any analysis by NTEE is subtly biased toward how the sector looked when its organizations were founded.

How to use NTEE without being misled

Five practical rules.

Aggregate at the major-group level for public claims. Letter-level groupings are considerably more robust than four-digit codes. "Human services" is a defensible aggregation; a specific four-digit activity code across thousands of organizations is not.

Treat it as one signal among several. Mission text, program descriptions in Form 990 Part III, grant purposes, and geography all say something about what an organization does. NTEE alone is the weakest of these.

Never use it to decide eligibility. Foundations that filter applicants by NTEE code reject organizations doing exactly the work they fund, because those organizations were coded thirty years ago. If you use NTEE at intake, use it to route, never to exclude.

Expect the T group to distort things. Grantmaking classification is inconsistent enough that any analysis of intermediaries should be sanity-checked directly.

Say so when you use it. Analyses built on NTEE should state that they are, so readers can weigh it.

What to use instead, or alongside

Because NTEE describes origin better than current activity, most serious analysis now supplements it.

Program text from filings. Form 990 Part III describes program service accomplishments in the organization's own words, updated annually. It is far more current than an NTEE code and considerably messier — which is exactly the trade-off.

Grant purposes. What a funder says a grant was for often describes the recipient's work better than the recipient's own classification.

Semantic similarity. Comparing the actual text of missions and programs groups organizations by what they resemble rather than by which box they were assigned. This catches the housing-support organization still coded under recreation, because its text reads like housing support.

This is the approach Plinth's dataset uses for cause and cluster analysis: NTEE where it is reliable, program text and embeddings where it is not, across the full e-filed universe of 205,036 grantmakers and 17,896,418 grants for fiscal years 2017 to 2025. It is also why look-alike sets sometimes surface organizations whose formal classification looks unrelated — the text says otherwise. Searching any organization is free without an account.

None of this makes NTEE obsolete. A stable, official, universally applied code is valuable precisely because everyone uses it, and comparability across studies has real worth. The right posture is to use it knowingly.

The major groups, briefly

Knowing roughly what each letter covers makes filings and datasets much faster to read. The 26 groups, in their conventional order:

Two of these deserve attention when reading data. T is where grantmaking foundations themselves sit, so any analysis that does not separate T will double-count money moving between institutions. And Z — unknown — is not empty; organizations end up there through incomplete records, and a dataset with a large Z population is telling you something about its own coverage.

The P group (human services) is by far the largest by organization count in most cuts of the data, which is one reason "human services" aggregations can look dominant while covering enormously varied work.

For nonprofits: check your own code

Worth ten minutes.

1. Look it up in the IRS Business Master File or via Tax Exempt Organization Search.
2. Ask whether it describes what you do now. Not what you were founded to do.
3. If it is materially wrong, consider correcting it. A misclassified organization is harder to find in every funder database that filters by cause — and many do.
4. Meanwhile, fix your Part III. Program service descriptions are updated annually, are read by researchers and increasingly by automated systems, and are entirely within your control. If your NTEE code is wrong, clear program text is the fastest available correction.

That last point generalizes. As more funder research runs on text rather than on codes, the words an organization uses to describe itself in its own filing matter more each year than the letter it was assigned at birth.

For funders: classifying your own portfolio

If you are analyzing your grantmaking by cause, you will hit the same problems in reverse.

Using grantees' NTEE codes gives you a portfolio picture skewed by their formation histories. Two mitigations:

Classify by grant purpose, not grantee identity. A grant to a multi-purpose organization for a specific housing program is housing funding, whatever the grantee's code says. This requires writing meaningful grant purposes at the point of decision — which, as covered in What Your Form 990-PF Reveals, is the highest-leverage data-quality habit available to a foundation.

Maintain your own taxonomy, mapped to NTEE. Most foundations think in categories that do not match NTEE's. Keeping your own and mapping to NTEE for external comparison gives you both internal coherence and comparability.

Where software fits

Classifying grants by cause is only possible if cause is captured at the point of decision. Reconstructing it later from grantee names and codes is where the distortion enters.

Tools like Plinth capture purpose and category as part of the grant workflow, so portfolio insights reports cause mix from what you recorded rather than from a third-party code assigned to your grantee decades ago. Applied consistently, this also improves the Part XV descriptions in your own filing — which is what every external analyst then uses to describe your foundation.

Why classification is getting more consequential

For most of NTEE's history, a wrong code was a minor statistical inconvenience. That is changing, for a reason worth naming.

Increasingly, the first pass over nonprofit data is made by software rather than a person — funder databases filtering by cause, matching tools proposing funder–grantee pairs, and answer engines summarizing "who funds X in Y." Each of those runs on structured fields first and text second, because structured fields are cheap to query.

The consequence is that a stale code no longer just misfiles you in a research report. It removes you from consideration before any human sees your name, and neither you nor the funder ever learns it happened. False negatives in classification are completely invisible to both sides.

Two implications follow.

For nonprofits: the code matters more than it did, and so does everything else machine-readable about you — Part III program descriptions, mission text, the specificity of how you describe your work. These are the fields that let a text-based system find you when the code fails.

For funders: any process that filters on classification should be audited for what it excludes. A useful exercise is to take ten organizations you know do work you fund, and check whether your own filters would have surfaced them. Foundations that run this test are usually surprised, and the surprise is always in the same direction.

None of this is an argument against classification. It is an argument for treating it as a lossy index into the real thing, rather than as the thing itself.

Frequently asked questions

What does NTEE stand for?

The National Taxonomy of Exempt Entities, a descriptive classification system for US tax-exempt organizations used by the IRS and the National Center for Charitable Statistics.

How many NTEE major groups are there?

Twenty-six, lettered A through Z, which NCCS rolls up into ten broad categories. Both figures describe the same system at different levels.

Where do I find an organization's NTEE code?

In the IRS Business Master File, via Tax Exempt Organization Search, or on most nonprofit data platforms, which read it from the same source.

Does my NTEE code affect my tax status?

No. NTEE is descriptive and used for statistical purposes. Tax treatment is determined by your 501(c) subsection and foundation classification.

Can an NTEE code be changed?

Yes, by request to the IRS, though the process is not instant and there is little enforcement pressure to keep codes current — which is why so many are outdated.

Why do funders sometimes miss organizations that fit their criteria?

Frequently because of stale NTEE codes. A database filtered by cause code will not surface an organization coded for work it stopped doing years ago.

Should foundations filter applicants by NTEE code?

No. Use it to route or prioritize, never to exclude, because false negatives are common and invisible — you never see the organizations the filter rejected.

Is NTEE used outside the United States?

No. It is a US taxonomy tied to the IRS exempt organization universe. Other countries use their own classifications, which is one reason cross-border comparisons of nonprofit sectors are harder than they look.

What is NTEE-V2?

A restructured format of the same taxonomy designed for analytics, which separates organization type from activity and exposes the higher-level industry code directly.

Recommended next pages

• What Your Form 990-PF Reveals — why grant purpose text matters so much
• Foundation Portfolio Analysis — classifying your own grantmaking
• How to Research a Foundation — researching past classification limits
• Co-Funder Networks — grouping funders by behavior rather than category
• How to Benchmark Your Foundation — cause mix as a benchmark axis

Last updated: August 2026