Methodology

How PeerBasis builds comparability data.

Version 1.2 · Updated July 18, 2026 · IRS Form 990 e-file through the 2025 processing year · BEA RPP 2024 · 2025-dollar comparison basis

PeerBasis produces descriptive compensation comparability data that a nonprofit board may consider as one input to its own decision under Internal Revenue Code section 4958. This page distinguishes filed values from derived values and documents the peer-selection and calculation rules. A report does not determine reasonableness or establish that a board has met the requirements for a rebuttable presumption.

01Data sources

PeerBasis's reported inputs and economic adjustment series come from public IRS, BLS, and BEA data. There is no proprietary compensation survey or self-reported panel; role classification, peer selection, adjustments, percentiles, and ranks are derived by PeerBasis and labeled as such.

IRS Form 990 e-file dataset: The electronic annual returns nonprofits file with the IRS, released as machine-readable XML. Compensation figures are read from Form 990, Part VII, Section A (officers, directors, trustees, key employees, and the highest-compensated employees), and organization financials from the return's summary.
IRS Business Master File (BMF): The IRS register of exempt organizations, used to attach each organization's National Taxonomy of Exempt Entities (NTEE) code and state for sector and geographic matching.

Each peer row carries an IRS filing identifier and links to a public organization record. Users should verify the applicable source filing. Price-level adjustments, inflation indexing, role classifications, peer selection, percentiles, and ranks are derived rather than filed values.

02The benchmarking figure

PeerBasis's default benchmark is Form 990 Part VII, Section A, column D + column F. Column D is reportable compensation from the filing organization. Column F is other compensation from the filing organization and related organizations, so D + F is a hybrid measure and must not be read as compensation solely from the filing organization. Column E, reportable compensation from related organizations, is displayed separately but excluded from the default benchmark. The report's sensitivity table also shows D-only and D + E + F alternatives when the source fields are available.

PeerBasis attempts to identify each organization's chief executive from the reported title (for example, executive director, CEO, or president serving as chief executive). Where no chief-executive title is identified, the highest-compensated eligible non-institutional individual is used as a fallback. A fallback may not be the chief executive, so it is labeled for board review rather than presented as an equivalent role.

03Exclusions

The default benchmark applies these exclusions and disclosures:

Related-organization compensation (column E). Column E is excluded from the default D + F figure and displayed separately. Because column F itself can include related-organization amounts, the default does not fully isolate filing-organization compensation.
Institutional trustees. Rows where the "person" is an organization (for example, a bank serving as trustee) are excluded, since they do not represent individual executive compensation.

04Peer-set selection

The matcher uses three observable structural dimensions as proxies for similarity. Whether the resulting organizations are sufficiently similarly situated for a particular decision remains a board judgment:

Sector: Same NTEE classification — first by specific sector code, then, if needed, by major group.
Filing size: A size-adaptive band uses positive total revenue by default. When expenses are positive, a revenue measure below 10% of expenses is treated as de minimis because it would create a token-size band; the exact 10% boundary remains eligible. If total revenue is nonpositive or de minimis, the matcher tries Form 990 gross receipts under the same rule, then total expenses and total assets. The chosen measure is applied to both the subject and every candidate, including the same de-minimis eligibility test for a candidate's revenue or gross receipts, and is named in the report. Bands are 0.67×–1.50× below $1M; 0.71×–1.40× from $1M to below $10M; 0.77×–1.30× from $10M to below $100M; and 0.80×–1.25× at $100M and above. If fewer than five peers remain, one documented widening to 0.50×–2.00× is attempted. If no usable measure exists, size matching is marked unavailable rather than constructing a zero or reversed band.
Geography: When the target has both a minor NTEE code and state, the first rung is that minor sector in the same state; when it does not reach the selection target, the matcher evaluates nationwide tiers. The report records the actual scope as same-state, nationwide, or unavailable.

To seek a more stable distribution, the criteria broaden in a fixed, documented order. The report always states the exact tier used.

Same NTEE minor sector and state, within the filing-size band (when both fields are present).
Same NTEE sector, nationwide, within the filing-size band.
Same NTEE major group, nationwide, within the filing-size band.

Within the selected tier, qualifying organizations are ordered by absolute distance on the disclosed filing-size measure. A runaway guard limits a result to the 2,000 closest rows; when that guard or disclosed user exclusions reduce the set, the report states the number included and the total qualifying count. An unlocked report lists every peer included in its calculations, including reported title and role-match status.

Price-level and comparison-year adjustments. For a nationwide tier, each peer with supported state data is restated at the subject state's consumer price level using BEA Regional Price Parities (RPP, all items, by state, 2024). RPP measures consumer price levels; it is not a local wage or cost-of-labor series. Peer filing figures and a last-reported subject figure with supported filing years are then indexed to a single 2025-dollar comparison basis using CPI-U annual averages; 2025 is the latest complete annual CPI-U year embedded in this version. Peer rows outside the embedded CPI year table are excluded from the comparison, and a last-reported subject amount with an unsupported year is not assigned a rank. A custom proposal entered by the user is treated as already expressed in 2025 dollars and is not CPI-adjusted, so an amount from another year must be restated before entry. If either state lacks an RPP, the interstate adjustment is a no-op for that row. Same-state peers need no interstate adjustment.

Sample size. For organizations with under $1 million in annual gross receipts, 26 CFR 53.4958-6 describes a special rule involving data from three comparable organizations. PeerBasis applies a separate product floor of five qualifying peers for every downloadable board report. A one-to-four-peer distribution may remain visible in the free preview, but checkout, PDF generation, and the board-minutes worksheet are disabled. A numeric minimum alone does not establish that a peer set is appropriate.

05Percentiles & interpretation

From the selected peers' adjusted D + F figures, PeerBasis computes the 10th, 25th, 50th (median), 75th, and 90th distribution percentiles using linear interpolation between sorted values. The subject's empirical rank is calculated differently: it is the percentage of selected peers with a strictly lower adjusted figure, rounded to the nearest whole percentage point. Tied peer values are not counted as lower.

Each report discloses the peer count, compensation and filing-year ranges, selected geography tier, subject and peer titles, direct-title versus fallback role status, column E flags, outlier counts, and—when the subject is a last-reported figure—a sensitivity table for D + F, D only, D + E + F, adjusted, and as-reported definitions.

Percentiles and ranks describe a position within the selected data. No percentile threshold automatically makes compensation reasonable or unreasonable, and PeerBasis does not make that determination.

06Legal basis: the rebuttable presumption

Under IRC 4958, a qualifying compensation arrangement may be presumed reasonable — shifting the burden of proof to the IRS — when the applicable requirements of 26 CFR 53.4958-6 are met, including:

Approval in advance by an authorized body composed of individuals without a conflict of interest.
Reliance on appropriate comparability data prior to the determination.
Adequate, concurrent documentation of the basis for the determination.

A PeerBasis report offers a structured source of comparability data that a board may evaluate for the second requirement and a documentation worksheet with prompts for the third. It does not establish that the data are appropriate for a particular decision, verify conflicts or advance approval, determine reasonableness, or establish the presumption. Form 990, Part VI, Line 15 asks filers whether they used a review and approval process for certain officials.

07Limitations

Price levels are state-level. The nonprofit matcher uses 2024 state Regional Price Parities, not metropolitan-area indices, so within-state variation is not captured. RPP measures consumer prices, not local wages or cost of labor.
Reporting periods vary. Supported filing-year amounts are indexed to 2025 comparison dollars with CPI-U, but organizations still file on different fiscal-year end dates, so comparables may reflect partly different twelve-month periods.
Titles are free text. Roles are normalized from free-text 990 titles. The most common titles map through a reviewable lookup table (labeled by a local large language model); the long tail can be assigned through an embedding nearest-neighbor process, with a rule-based fallback. The current selector uses compensation, reported hours, and title together to reduce—but cannot eliminate—the risk that a top-earning clinician or a part-time board officer is mistaken for the chief executive.

The report shows every included peer's reported title and flags a highest-paid fallback rather than presenting it as a direct match. In the June 2026 LLM-assisted validation sample, 28 of 28 direct selections checked matched the intended role, while only 7 of 15 non-ambiguous chief-executive fallbacks were judged to be a genuine organizational leader. The sample used one independent model labeler and is too small to establish population accuracy; every fallback requires board review.
Coverage is incomplete. The current database is a latest-filing-per-EIN snapshot of electronically filed full Forms 990, not a complete census of every exempt organization or every filing year. In the audited build, NTEE was present for 274,869 of 391,386 organizations (70.2%); organizations without a usable code cannot be sector-matched.
Structural comparability is limited. NTEE, state, revenue, reported title, and filing year do not capture every relevant fact, including duties, experience, hours, performance, organization complexity, controlled entities, or unusual one-time pay.
Not advice. A PeerBasis report is comparability data and documentation support. It is not legal or tax advice, and boards should consult their own counsel.

08Provenance & citation

Reports are dated and record the peer-selection tier, included count, filing-year range, comparison-year basis, and raw versus comparison figures so the computation can be reviewed. Organization links resolve to public organization records; users should identify and verify the applicable source filing. To reference this methodology, cite: "PeerBasis Methodology, Version 1.2," peerbasis.org/methodology, updated July 18, 2026.

09Glossary

Plain-language definitions of the terms used throughout PeerBasis.

IRC 4958: Internal Revenue Code section 4958 imposes excise taxes on excess-benefit transactions involving applicable tax-exempt organizations. A disqualified person can owe tax on the excess benefit; an organization manager tax requires knowing and willful participation that was not due to reasonable cause.
Rebuttable presumption of reasonableness: A regulatory presumption under 26 CFR 53.4958-6 that may apply when compensation is (1) approved in advance by an authorized body free of conflicts of interest, (2) supported by appropriate comparability data the body relied on, and (3) documented concurrently in written or electronic records. The report alone does not establish it.
Comparability data: Data on compensation paid for similar services by similarly situated organizations — comparable in mission, budget, and location. PeerBasis assembles it from IRS Form 990 filings.
IRS Form 990: The annual information return most tax-exempt organizations file with the IRS. Part VII, Section A and Schedule J report officer, director, trustee, and key-employee compensation. The IRS releases the e-filed returns as machine-readable XML.
PeerBasis D + F comparison figure: PeerBasis's default hybrid measure: reportable compensation from the filing organization (Form 990 Part VII column D) plus column F other compensation. Column F may include amounts from the filing organization and related organizations. Column E is excluded from the default and displayed separately; D + F therefore must not be labeled compensation solely from the filing organization.

Build a comparability report