0003: XBRL/LOGERx transition: no structural change in EDR’s compiled series

Context

SRC-06 requires the XBRL/LOGERx reporting-system change (fiscal years ending on or after September 1, 2022) to be assessed for structural impact on the series. .planning/research/PITFALLS.md Pitfall 4 documents the underlying legislated change: Florida mandated XBRL reporting for local governments, submitted through the Florida Open Financial Statement System (LOGERx, at DFS), for fiscal years ending on or after September 1, 2022 — a structurally different submission system from the pre-2022 filing process. .planning/phases/01-source-data-verification/01-CONTEXT.md D-14 decides how this is assessed: against the EDR-compiled files already on disk, not by pulling a raw filing — the question is whether the transition changes the structure of the series TaxpayerScope actually consumes (EDR’s compiled per-county data), and it is answered directly: does the account-code set, column layout, fund-type labeling, or granularity shift at the FY2021-2022 boundary across counties, in a way that stands out from ordinary year-to-year change? A systemic break in the same year across many counties would be the signal; ordinary account-code churn happening every year regardless of the reporting system would not be.

Plan 01-03 produced the quantified evidence this record answers from (data/audit/structure-year-diff.csv); this record closes SRC-06.

Decision

The XBRL/LOGERx transition does not change the structure of EDR’s compiled series. Answered directly from the archived files, not from agency documentation: at the FY2021-2022 boundary — the fiscal-year boundary the September 1, 2022 XBRL mandate falls inside — zero fund-type columns were added or removed for any of the 67 governments (data/audit/structure-year-diff.csv: Custodial added for 0/67 government(s) at this boundary). The raw signature_changed fraction at this boundary is 100% (every one of 67 governments shows some structural difference between its FY2021 and FY2022 sheets), but that figure does not stand out — the median background change rate across every other adjacent-year boundary in the ten-year window is 97.8%, nearly identical. Ordinary account-code churn (codes added or dropped from one year to the next) happens at almost every boundary in this corpus regardless of which reporting system produced the filing; FY2021-2022 is statistically unremarkable against that background.

This stands in direct contrast to the FY2020-21 GASB 84 boundary (ADR-0002), where the same signature_changed measure is also 100%, but the specific signal — a new fund-type column being added — is 100% of governments at 2020-2021 versus 0% of governments at 2021-2022. The GASB 84 boundary is a genuine systemic structural break (every government gains an entirely new Custodial column, the same year, uniformly); the XBRL/LOGERx boundary is not (no government gains or loses a fund-type column at that boundary at all). Stating the conclusion as this comparison — background rate versus targeted signal, at two adjacent boundaries in the same corpus — rather than as a bare assertion is the point of this record.

Raw XBRL and LOGERx access is fallback-only and deliberately unbuilt. No code in this project parses a raw XBRL instance document or queries the LOGERx submission system directly. The condition that would change this: the EDR-compiled files proving insufficient for a specific gap, correction, or fiscal year — for example, a fiscal year EDR has not yet compiled into its per-county Excel file, or a discrepancy between a compiled figure and a government’s own published audited financials that the compiled file’s structure cannot resolve.

Rationale

EDR’s compiled files are sourced from DFS Annual Financial Report submissions regardless of which underlying filing system produced them — the pre-2022 submission process and LOGERx/XBRL alike feed the same EDR compilation pipeline that produces the per-county Excel files this project consumes. This is exactly why a filing-system change can leave the compiled series structurally unchanged: EDR’s own compilation step is the layer that normalizes across whichever submission mechanism produced a given fiscal year’s data, and that normalization is what this project’s source-file choice (D-14: consume the compiled files, not a raw filing) inherits for free.

Rejected alternative: parse raw XBRL instance documents or query the LOGERx submission system directly for every fiscal year, rather than trusting EDR’s compiled Excel files. Choosing this would have cost a second, structurally different parser (raw XBRL instance documents look nothing like EDR’s compiled per-county Excel sheets) built and maintained speculatively, ahead of any demonstrated gap in the EDR-compiled files this project actually consumes. Building that path now is exactly the anti-pattern CLAUDE.md’s “What NOT to Use” table and .planning/research/PITFALLS.md Pitfall 1 both warn against: committing to adapter/parser structure before the actual need is known. The evidence above shows the EDR-compiled files already absorb the XBRL/LOGERx transition without a structural break, which removes the reason to build the raw-XBRL path speculatively.

What a future reader should check first if the transition is suspected to matter after all: re-run pipeline/exploration/structure_diff.py --boundary-year 2021-2022 (or its Phase 3 successor) against the then-current corpus and compare the fund_types_added/fund_types_removed columns in the output to the FY2021-2022 rows already committed in data/audit/structure-year-diff.csv — if a later fiscal year’s post-XBRL filing produces a fund- type column addition, formula-structure change, or account-code-granularity shift that this FY2021-2022 boundary did not, that is the concrete, checkable signal that would reopen this assessment, not a general suspicion about the filing system.

Source Evidence

Every figure below is quoted from data/audit/structure-year-diff.csv (committed, 1206 rows, Plan 01-03) via pipeline/exploration/structure_diff.py --boundary-year <year> re-run this task:

2020-2021 boundary (GASB 84, ADR-0002’s mechanism, quoted here as contrast):

Boundary 2020-2021: 67/67 governments changed structure (fraction 100.0%)
  Custodial added for 67/67 government(s) at this boundary (fraction 100.0%)
  Most common fund-types added: [('Private Purpose', 67), ('Custodial', 67)]
  Most common fund-types removed: []
Background (median fraction changed across all other boundaries): 97.8%

2021-2022 boundary (the XBRL/LOGERx transition boundary):

Boundary 2021-2022: 67/67 governments changed structure (fraction 100.0%)
  Custodial added for 0/67 government(s) at this boundary (fraction 0.0%)
  Most common fund-types added: []
  Most common fund-types removed: []
Background (median fraction changed across all other boundaries): 97.8%

The fraction 100.0% figure in both blocks is the raw signature_changed rate (any difference in fund-type label set, total-column label, or account-code set between the two adjacent years’ rows in data/audit/structural-audit.csv) — this is the “boundary change rate” the plan’s own <verify> gate requires quoting, and it is identical at both boundaries, which is precisely why it alone cannot distinguish a genuine structural break from ordinary churn. The distinguishing signal is the second line: whether a fund-type column itself (Custodial, in this case) was added, which is 100% at 2020-2021 and 0% at 2021-2022. The median background rate (97.8%) is the point of comparison: FY2021-2022’s raw 100% change rate is statistically indistinguishable from “an ordinary year” in this ten-year corpus, while FY2020-21’s fund-type-column-added signal has no counterpart anywhere else in the window.

Methodology Version

This decision takes effect in methodology_version: 1. It has no prior version to supersede — this is the first record of the XBRL/LOGERx structural assessment.

Revision (2026-09-24)

This revision supersedes methodology_version: 1 with methodology_version: 2. The decision recorded above is carried forward unchanged into methodology_version 2; the changes this version-2 coordinated bump makes are recorded in 0001-total-expenditures-definition.md, 0005-population-and-cpi-series.md and 0008-total-revenue-definition.md. This revision recorded by Task 3, 03-06-PLAN.md.

Revision (2026-09-24)

This revision supersedes methodology_version: 2 with methodology_version: 3 (ADR-0009, 04-01-PLAN.md). The decision recorded above is carried forward unchanged into methodology_version 3; the changes this version-3 coordinated bump makes are recorded in 0009-category-and-revenue-classification.md. This revision recorded by Task 3, 04-01-PLAN.md.

Revision (2026-09-25)

This revision supersedes methodology_version: 3 with methodology_version: 4 (ADR-0010, 05-01-PLAN.md). The decision recorded above is carried forward unchanged into methodology_version 4; the changes this version-4 coordinated bump makes are recorded in 0010-statewide-publication-rules.md. This revision recorded by Task 3, 05-01-PLAN.md.