Methodology
This tool builds compensation benchmarks from publicly filed IRS Form 990 data. Here's how the data gets from the IRS to your report.
Data Source
All compensation data in this tool comes from IRS Form 990 returns filed electronically by tax-exempt organizations. These filings are public records, available through the IRS Exempt Organizations Business Master File and the GivingTuesday Data Commons.
We index filings from tax years 2021–2025, covering both Form 990 (standard) and Form 990-EZ (simplified) filings. Form 990-N (postcard) and 990-T (unrelated business income) filings are excluded as they do not contain compensation data. Data for 2024 and 2025 is available but incomplete — see Limitations below for details.
Compensation Data
Compensation figures are extracted from Part VII, Section A of Form 990, which reports compensation for officers, directors, trustees, key employees, and highest compensated employees.
Total compensation is calculated as the sum of three components reported on the return:
- Reportable compensation from the organization — W-2 or 1099-MISC income paid directly by the filing organization
- Reportable compensation from related organizations — compensation from affiliates, subsidiaries, or parent organizations
- Other compensation — deferred compensation, nontaxable benefits, and other forms of non-reportable income
For organizations that file Schedule J (Compensation Information for Certain Officers, Directors, Trustees, Key Employees, and Highest Compensated Employees), we also capture the detailed breakdown: base compensation, bonus/incentive pay, other reportable income, retirement and deferred compensation, and nontaxable benefits.
Role Normalization
Nonprofits report job titles in freeform text, leading to hundreds of variations for the same role (“CEO,” “Chief Executive Officer,” “Executive Director,” “President & CEO,” etc.). To enable meaningful comparison, we normalize these titles into eight canonical roles:
The normalization engine uses a curated mapping of 160+ known title variants supplemented by pattern-matching rules for less common phrasings. Titles that cannot be confidently mapped are excluded from role-based benchmarking.
Building a Comparability Set
A comparability set (or “peer group”) is the foundation of a defensible compensation analysis. The IRS expects organizations to demonstrate that executive compensation is reasonable by comparing against similarly situated organizations. Our builder lets you define peers using four dimensions:
- Sector (NTEE Code)
- Organizations are classified by mission area using the National Taxonomy of Exempt Entities. NTEE codes are sourced from the IRS Business Master File, the canonical registry for exempt organization classifications.
- Budget Size
- Total annual expenses serve as the proxy for organizational size. Budget tiers are aligned with Candid/GuideStar compensation study breakpoints, from under $250K to $50M+. Custom ranges are also supported.
- Geography
- Filter nationally or by individual states. Geographic filtering uses the organization's state of domicile as reported on the 990.
- Schedule
- Full-time (30+ hours per week) or part-time, based on average hours reported on the return.
When multiple filings exist for the same organization, only the most recent tax year is included to avoid double-counting.
Statistical Methods
Compensation statistics are computed from the filtered comparability set. We report the following measures:
- Median — the midpoint of the distribution (50th percentile). Less sensitive to outliers than the mean, making it the most reliable single benchmark.
- Mean — the arithmetic average. Useful for context but can be skewed by a few very high or very low values.
- Percentiles (10th, 25th, 75th, 90th) — computed using linear interpolation between the two nearest ranked values. These show where a specific compensation figure falls within the distribution.
- Range — the minimum and maximum total compensation in the set.
The distribution chart uses dynamic histogram bucketing, rounded to the nearest $5,000 increment, to show the shape of the compensation distribution across the peer set.
Data Quality & Outlier Handling
Raw 990 data contains errors, edge cases, and outliers that would distort benchmarks if left unchecked. We apply the following filters transparently:
- Zero compensation excluded. Records with $0 total compensation (common for unpaid board members or data entry errors) are excluded from all benchmarking results.
- Compensation capped at $10M. Figures exceeding $10,000,000 are excluded as data errors. The highest-paid nonprofit executives (major health systems) rarely exceed $5–7M in total compensation.
- Negative financial values clamped. IRS instructions specify that negative amounts should be reported as zero. Any negative total expenses or revenue values encountered during processing are treated as zero.
- Most recent filing only. When multiple filings exist for the same organization and role, only the most recent tax year is included, preventing double-counting.
We prefer the median over the mean as the primary benchmark because it is inherently resistant to outliers — a practice consistent with the Candid Nonprofit Compensation Report and other industry-standard references.
Limitations & Caveats
- Filing lag. Form 990 is due on the 15th day of the 5th month after the fiscal year ends, with extensions available. Most filings appear 12–24 months after the period they cover. The data here reflects the most recent available filings, not current compensation.
- 2024–2025 data is incomplete. As of early 2026, most organizations with calendar fiscal years have not yet filed their 2024 returns (still within extension windows). Fiscal-year filers with year-ends before June 2025 may already be available. Both years are indexed and will grow as filings arrive — but peer sets anchored to these years will be smaller and less representative until coverage fills in.
- Self-reported data. Compensation figures are reported by the filing organization. While 990s are signed under penalty of perjury, errors and inconsistencies do occur.
- E-file only. This tool indexes electronically filed returns. Organizations that file on paper are not included.
- Title normalization is imperfect. Some titles are ambiguous (“Director” could mean board member or department head). Our mapping is conservative — when in doubt, a title is excluded rather than mis-classified.
- NTEE codes are not always current. The Business Master File is the best available source, but some organizations may have outdated or missing classifications. When sector filters are applied, organizations with no NTEE code on file are included to avoid artificially narrowing the set.
- Not legal or tax advice. This tool is for informational purposes. Compensation decisions should be made in consultation with qualified advisors and in compliance with IRS intermediate sanctions regulations (Section 4958).
IRS Rebuttable Presumption of Reasonableness
Under IRS regulations, a nonprofit can establish a rebuttable presumption of reasonableness for executive compensation if three conditions are met:
- Compensation is approved by an authorized body composed entirely of individuals with no conflict of interest
- The authorized body obtained and relied upon appropriate comparability data before making its determination
- The authorized body adequately documented the basis for its determination concurrently with making that determination
This tool helps satisfy the second requirement by providing comparability data drawn from actual IRS filings. The generated report documents the peer group criteria and resulting compensation benchmarks, supporting the third requirement.
Data sourced from publicly filed IRS Form 990 returns via the GivingTuesday Data Commons and IRS Exempt Organizations Business Master File. For informational purposes only — not legal or tax advice.
