Nonprofit Data and Reporting: Where the Numbers Come From

Almost every statistic quoted about the US nonprofit sector traces back to a small number of sources, all of them free and public. Knowing which is which saves you from citing a figure that turns out to be a blog quoting a blog.

The sources, and what each one holds

Note the currency column, because it is where most misuse happens. Giving figures describe the year just ended. IRS financial statistics and BLS employment data currently describe 2022. Anyone quoting a current-year employment or financial figure for the sector is estimating, whether or not they say so.

Nonprofit employment data specifically

The authoritative series comes from the Bureau of Labor Statistics Quarterly Census of Employment and Wages, which counts jobs at establishments filing payroll returns. In 2022 that was 12.8 million jobs at more than 300,000 nonprofit establishments, 9.9% of private sector employment.

The analysis that made this data usable was the Johns Hopkins Nonprofit Economic Data Project, run from 1999 at the Johns Hopkins Center for Civil Society Studies under Lester Salamon. It extracted nonprofit employment and wage figures from the QCEW and published them nationally and by state, which nobody else was doing.

If you are looking for that project now, the work has substantially moved. George Mason University’s Schar School runs a Nonprofit Employment Data Project building on the Johns Hopkins effort, and has published more recent nonprofit employment reports. Start there for current figures, and treat older Johns Hopkins publications as the historical series.

One caution that applies to all of it: household surveys find more nonprofit workers than establishment surveys do, because they count people who say they work for a nonprofit rather than jobs at employers who file. Both are correct and they answer different questions.

Finding data on organizations like yours

This is the most useful research available to a nonprofit and almost nobody does it.

Every organization filing a full Form 990 or a 990-EZ discloses its revenue by category, its expenses by function, its largest contractors and the compensation of its highest paid people. Three consecutive years must be available for public inspection, and in practice they are all online free.

Pick five organizations doing similar work at a similar size, pull three years each, and you will learn more about what is normal for you than any national benchmark will show. The foundation returns are better still: a 990-PF is a list of exactly who funds this kind of work and how much they give, which is a better prospect list than most paid databases produce.

Why sector averages mislead

Three reasons, and they compound.

The dataset contains giants. Hospitals and universities sit in the same figures as an all-volunteer food bank. Most published statistics are means, and means in this sector are dominated by the largest filers. Medians are far more useful and far less often published.

Most nonprofits are invisible in it. Around 1.97 million organizations are registered and only about 300,000 have any paid employees. Any benchmark drawn from employers describes a small minority of the sector.

Subsector matters more than sector. An arts organization and a human services organization face different donor bases, funders and seasonality. Sector-wide donor retention of 43.3% is a conversation starter, not a target.

Reporting your own numbers

Two audiences, and they want different things.

Your board needs actual against budget with an explanation of any material variance, unrestricted net assets, and how many months of expenses those would cover. A statement with no comparison tells a director what happened and not whether it was expected.

Your funders want the measures you promised, in the words you promised them, so they can see at a glance whether it happened. Report against your own application rather than inventing a new frame, and say plainly where something did not work.

Choose measures you can actually gather with the staff you have. An outcome framework requiring a full-time evaluator, at an organization with one part-time coordinator, will not be collected, and the funder who was promised the data will notice.

Building a simple dashboard

Most small organizations either track nothing or track everything. A dozen numbers, reviewed monthly, is the useful middle.

The donor count row is the one organizations most often lack, and it is where the sector’s real trend hides. Nationally, giving totals rose while donor numbers fell 3.6% in 2025. An organization tracking only income would read that as a good year.

Collecting outcome data you will actually use

Four questions, answered before you collect anything.

What changes for the person? That is the outcome. Sessions delivered is an activity.

What would you see if it changed? That is the indicator, and it has to be observable by someone who is already there.

Who records it, when, in what? If the answer is nobody in particular, at some point, somewhere, it will not be recorded.

What would count as working? Set the target before you see the data, not after.

Two or three outcomes tracked properly beats nine tracked aspirationally, and it is the difference between an organization that learns and one that reports.

Citing figures properly

Three habits that keep you out of trouble, on a website, in a grant application or in an annual report.

Name the body that published it, not the site you read it on. Attach the year the data describes, which is not the year it was published. And where you cannot confirm a figure at its source, say so rather than repeating it, because a number that cannot be traced is a number somebody will eventually ask you about.

Questions people ask

Where does nonprofit sector data come from?

A small number of sources, all free. The IRS publishes Statistics of Income compiled from Forms 990, plus Tax Exempt Organization Search and machine readable bulk filings. The Bureau of Labor Statistics publishes nonprofit employment and wages from the Quarterly Census of Employment and Wages.

Giving USA, researched by the Indiana University Lilly Family School of Philanthropy, publishes annual charitable giving totals. The Fundraising Effectiveness Project publishes donor counts and retention quarterly.

ProPublica Nonprofit Explorer and Candid make individual organizations' filings searchable.

Almost every statistic you will read about the sector traces back to one of these.

What is the Johns Hopkins Nonprofit Economic Data Project?

A research project begun in 1999 at the Johns Hopkins Center for Civil Society Studies, led by Lester Salamon, which extracted nonprofit employment and wage data from the Bureau of Labor Statistics Quarterly Census of Employment and Wages and published it nationally and by state.

It made nonprofit employment visible in a way no other source did, and much of what is now quoted about the sector as an employer originates in it.

The work has substantially moved on. George Mason University's Schar School now runs a Nonprofit Employment Data Project building on the Johns Hopkins effort, with more recent employment reports. Use that for current figures and treat the older Johns Hopkins publications as the historical series.

How current is nonprofit data?

Less current than it looks, and unevenly so.

Giving USA is the freshest, published in June for the year just ended. Fundraising Effectiveness Project data is quarterly.

IRS Statistics of Income currently covers tax year 2022, because it is compiled from returns filed months after those years ended. BLS nonprofit employment also stands at 2022 and is refreshed roughly every five years rather than annually.

Treat any current-year employment or financial figure for the sector as an estimate, and be suspicious of any source quoting one without a year attached.

How do I find financial data on a specific nonprofit?

ProPublica Nonprofit Explorer, free, which holds Form 990 filings and extracted financials and is searchable by name, location or employer identification number.

IRS Tax Exempt Organization Search is the authoritative record for status: whether the organization is exempt, under which subsection, whether contributions are deductible, and whether exemption has been revoked.

Candid also holds profiles and filings, with free and paid tiers.

Search on the employer identification number where you can. Names are not protected, similar names are common, and matching on name alone is how people end up looking at a different organization.

What can you learn from another nonprofit's Form 990?

A great deal, and it is public by law.

Revenue by category, so you can see their actual funding mix. Expenses split across programme, management and general, and fundraising. Compensation of officers, directors and the highest paid employees. Largest independent contractors. Governance answers, including whether they have a conflict of interest policy and how many directors are independent. And transactions with interested persons, on Schedule L.

Foundation returns, the 990-PF, go further: they list every grant made, with recipient and amount, which tells you exactly who funds work like yours and at what level.

Why are nonprofit averages misleading?

Three reasons that compound.

The dataset contains hospitals and universities alongside all-volunteer groups, and most published figures are means, which the largest filers dominate. Medians are far more useful and far less often published.

Most nonprofits barely appear. Around 1.97 million organizations are registered and only about 300,000 have paid employees, so any benchmark drawn from employers describes a small minority.

And subsector matters more than sector. An arts organization and a human services organization face different donors, funders and seasonality.

The useful comparison is five organizations of your size in your field, from their own filings.

What should we report to our board?

Actual against budget, with an explanation of any material variance. A statement without a comparison tells a director what happened, not whether it was expected.

Three numbers worth putting in front of them every time: unrestricted net assets, how many months of operating expenses those would cover, and any line materially over or under plan.

Add progress against the strategic and fundraising plans, reported by the person who owns each item rather than by the executive on everyone's behalf.

A board that receives statements and never discusses them is not exercising oversight, and duty of care applies whether or not directors read the papers.

How should we cite statistics in a grant application?

Name the body that published the figure rather than the website you read it on, and attach the year the data describes, which is not the year it was published.

Where you cannot confirm a figure at its original source, leave it out. A programme officer who checks a number and cannot find it has learned something about the rest of your application.

Pair national data with something local you know first hand. National statistics establish that the problem exists; what you know about your own area is what makes the application yours rather than a template.

This is reference information, not legal or tax advice. Rules vary by state and change over time. For a decision that carries real consequences, check the current text at irs.gov or your state registry, and talk to a nonprofit attorney or CPA.