Comparison Analysis Protocol

Version 1.0. Issued 13 September 2026; adopted 20 September 2026 (LOG-001 D-024, D-026). Status: adopted. Governs every cross-state file analysis the Index publishes: the national rollup, the state and class summaries on the site, the exports that place state files side by side, and any figure that counts across state files. Adopted on the owner's and principal's decisions of 13 September 2026.

1. Principle

A documentary record turns into a league table at the moment a reader can sort it by merit. Every rule below exists to make comparison possible and ranking impossible. The Index publishes distributions, counts, and named lists drawn from released state files; it publishes no ordering by merit, no composite score, and no characterization of any body beyond what the state file that names it says.

2. Inputs

A comparison is drawn only from released versions. No state file under notice, in draft, or withdrawn enters a comparison. Every comparison lists its constituent state files by identifier and version, and is re-issued as a new version whenever any constituent changes, under VER-002. A comparison carries the framework versions its constituents were coded under, and where they differ, section 3 decides what may be compared.

3. Comparability

Every cross-state file measure carries a comparability marker, computed and shown beside it:

Comparable. The measure has the same definition, the same reference year or cycle, and the same framework version across every constituent, and every constituent's frame is of the same type (exhaustive frame, or a stated reproducible sample under GOV-002).

Comparable with a stated caveat. One of those conditions differs and the difference is stated in words beside the measure, with the constituents affected named (for example California's stated sample beside exhaustive frames, or a file coded under FRM-004 v1.0 beside files coded under a later version).

Not comparable. The measure is shown per state file only, with no cross-state file count, because the definitions or the frames differ in a way that a caveat cannot carry.

The legacy files 01 through 10 enter no comparison on the fiscal dimension until their backfill is released, and enter comparisons on the other dimensions with the caveat that they predate the conventions at STD-002 v1.0.

4. Presentation

Results are shown as distributions: the count and the share of local education agencies at each coverage code within and across states, by agency type, against a stated denominator (the full frame and, where the Index distinguishes it, the readable subset). No mean, median, or composite of coverage codes is published, because the coverage code is a position on a published rubric and not a quantity. Default ordering of any table is alphabetical or by identifier; where a column is sortable, the page states that sort order is not a ranking; and no table, chart, or export carries the words best, worst, top, bottom, leader, or laggard. Charts show distributions and change over time; they do not show one state file against another on a scale of merit.

5. Named lists

A named list (every agency with a given coverage code, every state whose figures changed between cycles) is drawn only from state file rows, links to the rows it is drawn from, and says nothing about a body that its state file does not say. A body's route to comment on its appearance in a list is the comment route of the state file that names it under STD-006; a list needs no separate notice. The exception is a cross-instrument profile of one institution, which is a new document about one body and receives notice under STD-005 like a state file.

6. Constructs and interpretation

The cross-state file measures are the constructs at FRM-005 and nothing else. A new construct enters this protocol by a decision at LOG-001 before it is published. Interpretation appears only in the national rollup's analysis section, marked as interpretation, and never recommends anything. What a state does not publish or does not measure is reported as a finding in the Index's formulations and is not characterized.

7. Timing and versions

A national rollup published before every state file in the frame has a released state file is labeled provisional and carries the count ("N of 51"). The full national rollup is published at the completion of a baseline or cycle, under its own version and DOI, and is re-issued by minor version when a constituent changes and by major version at each cycle. The change between cycles is the finding the Index exists to produce and is reported as such, with the delta by coverage code, figure, or frame decision and direction.

8. Review and reproducibility

The national rollup is built by script from the release packages, and the script, its inputs by version, and its output are deposited with the national rollup's DOI so that anyone can rebuild it. The method advisory group at GOV-005 reviews the comparison method annually; before the first full national rollup an external methods reviewer reads the method and this protocol, and the reviewer's comments and the Index's responses are published with the national rollup.

9. What is never published

A ranking. A score that combines measures. A grade, a badge, a seal, a certificate, or a tier of merit. A statement that any state file complies or fails to comply with any law. A characterization of a body that its state file does not carry. A comparison that includes a state file under notice.