How to Benchmark a Nonprofit Executive Director Salary

The question a board is actually asking

When a nonprofit board asks whether its Executive Director is paid appropriately, it is not asking for a market average. It is asking a narrower and more answerable question: what do organizations that look like this one pay someone doing this job, and can we show our work?

That framing matters because the sector's compensation data is unusually good in one respect and unusually awkward in another. Every filing organization's pay is public through Form 990 Schedule J. But nonprofit organizations vary so widely — in budget, mission, staff size, and geography — that a raw sector average is close to meaningless.

The work is in peer selection, not in data access.

Why the title tells you very little

"Executive Director" is applied to the chief executive of a two-person advocacy group and to the chief executive of a large regional health or education organization. The same title covers roles with radically different scope. Some organizations use "President and CEO" for the identical job.

This means title-based matching fails immediately, in a way it does not for public-company CEOs. A peer set assembled by searching for the title will contain organizations that have nothing in common with yours.

Two variables do most of the explanatory work instead: budget band and mission area.

Budget band is the primary screen

Total annual expenses — reported on the 990 — is the nonprofit equivalent of the revenue band used to size public-company peer sets. It is the closest available proxy for organizational scope, and it correlates with executive pay more reliably than any other single figure on the form.

Bands should be wide enough to produce a usable peer count and narrow enough to be meaningful. An organization with $4 million in annual expenses is not comparable to one with $60 million, and treating them as peers because both are "mid-sized nonprofits" produces a range so wide it cannot support a decision.

The same logic that governs public-company size banding applies here, and it breaks down in the same place — at the small end, where the number of comparable filers thins out. The problem is described in detail for the corporate case in CEO salary at a $50 million revenue company, and the nonprofit version is structurally identical.

Mission area is the second screen

Organizations are classified by mission, and the classification is on the filing. A hospital system, a private foundation, a university, an arts organization, and a direct-service human-services agency operate under different funding models, different regulatory environments, and different labor markets for senior staff.

Mission area matters because it changes the competitive set for the role. An ED at a research-funded organization is competing for talent with a different pool than an ED at a membership association of the same budget size.

Geography is a third screen worth applying where the peer count allows it, though it is often the first one to relax when the set gets too small.

Normalize before you compare

Schedule J reports compensation in components, and organizations distribute pay across those components differently. Comparing base salary alone will systematically understate organizations that pay more through retirement contributions or benefits.

The normalization steps are the ones covered in the Schedule J reading guide: align tax years rather than filing dates, identify one-time events sitting in other reportable compensation, aggregate compensation paid across related organizations, and check column (F) so a deferral is not counted twice.

Only after those adjustments does a total from one organization mean the same thing as a total from another.

Reading a distribution rather than an average

Once you have a normalized peer set, the useful output is a distribution, not a single number. Percentile bands — the 25th, 50th, 75th, and 90th — tell a board something an average cannot: how tightly the peer set clusters, and where in the range the organization currently sits.

An average conceals its own spread. If four organizations pay near one figure and one pays far above it, the average describes none of them. The percentile view makes the shape visible, and the shape is usually what the board actually needs to see.

The same reasoning applies to executive pay generally, and is worked through in calculating your CEO total compensation percentile.

What the data cannot settle

Three limits are worth being explicit about with a board.

The filings lag. Form 990 data is typically available later than the equivalent corporate disclosure, so the most recent comparison you can build is not current-year.

Small mission areas thin out fast. Apply budget band, mission area, and geography simultaneously to a specialized field and you can arrive at three usable peers. Three is a data point, not a distribution, and the honest response is to widen a screen and say which one you widened.

The data describes what organizations paid, not what they should have paid. A peer set is evidence for a decision, not the decision itself. Reasonableness in the nonprofit context involves considerations — mission, stewardship of donor funds, board judgment — that no percentile band resolves.

A worked sequence

The process is short enough to state end to end.

Begin with your own organization's most recent filing, so you know precisely what you are positioning: the ED's total by component, the tax year, and the organization's total annual expenses.

Define the screens before looking at any peer. Budget band as a range around your own expense figure, mission area, and geography if the peer count supports it. Writing the screens down first is what stops the peer set from being assembled backward to reach a desired answer — which is the failure mode a board is most entitled to worry about.

Identify candidate organizations meeting the screens and pull each one's filing. Record the EIN, tax year, total expenses, and the Schedule J component breakdown for the chief executive.

Normalize. Align tax years, flag one-time amounts, aggregate related organizations, net out previously reported deferrals.

Compute the distribution and place your own figure in it. Then write down which screens you relaxed, if any, and why.

The whole exercise is a few hours of work against public documents. What it produces is not a better guess — it is a different category of object, one where every input can be checked by anyone who doubts the conclusion.

What boards do with the result

Nonprofit compensation decisions are typically documented in a way corporate decisions are not, because the organization is accountable to a public interest rather than to shareholders alone.

In practice that means the peer analysis is not just an input to the board's judgment; it is part of the record of how the judgment was made. Boards commonly want to show that they reviewed comparability data for similar positions at similarly situated organizations, and that they did so before setting compensation rather than to justify it afterward.

This has a practical consequence for how the analysis should be built. A file assembled from a paid survey the board cannot inspect is weaker documentation than a file assembled from named filings anyone can retrieve — even if the survey has a larger sample. Checkability is doing work here that sample size does not.

The same distinction applies to salary data at every level, and SalaryRange makes the case for staff pay in why crowd-sourced salary estimates are not documentation.

When the peer set is too small

Specialized mission areas in modest budget bands routinely yield too few peers, and the honest responses are limited.

You can widen the budget band and note that you did. You can relax geography, which is usually the least damaging screen to drop. You can broaden the mission area to an adjacent one, saying which and why. Or you can supplement Form 990 peers with public-company comparables for roles of similar scope, being explicit that you are crossing disclosure regimes and comparing cash to cash.

What you should not do is present a distribution computed from four organizations as though it carried the authority of forty. A range built on a thin set is still useful — it is simply a range with a stated limitation attached, and stating the limitation is what preserves the credibility of everything else in the file.

What a defensible file looks like

At the end of the exercise you should be able to hand someone a record showing: the screens you applied and why, the organizations that met them, each one's EIN and tax year, the normalized total for each, the adjustments you made to get there, and the resulting distribution.

That file is checkable. Anyone can pull the same filings and verify the figures, which is precisely what makes it useful in a room where the alternative is an unsourced number and a disagreement about whose intuition is better.

The Nonprofit ED Pay Comparables Guide & Form 990 Peer Workbook helps you assemble that file: where ED pay is disclosed, how to choose peers by mission type and budget, and a workbook that computes percentiles and Schedule J component medians over the peers you enter, each with an EIN and tax year. To bring each peer's components to one comparable total, the Nonprofit ED Total-Compensation Normalization Workbook handles the normalization step for each organization you enter.