0002: GASB 84 custodial-fund series: replicate the adjusted exclusion uniformly
Context
SRC-05 requires the FY2020-21 GASB 84 custodial-fund discontinuity to be characterized, and the
decision to use EDR’s adjusted or unadjusted series recorded with rationale.
.planning/research/PITFALLS.md Pitfall 3 documents the mechanism directly: the Florida
Department of Financial Services added a Custodial Fund column to the AFR starting FY2020-21,
in preparation for GASB Statement No. 84 (Fiduciary Activities). EDR’s own published statewide
files state that “account totals for FY 2020-21 and subsequent fiscal years may not be directly
comparable” to earlier years without adjustment, and EDR itself publishes both an “unadjusted” and
an “adjusted” (custodial-excluded) statewide total for this reason. .planning/phases/01-source-data-verification/01-CONTEXT.md
D-12 decides how TaxpayerScope answers this: replicate the exclusion ourselves, with one uniform
rule, rather than depend on whether EDR happened to publish an adjusted figure for a given
per-county file — and validate the replicated roll-up against EDR’s own published adjusted
statewide totals as an independent confirmation.
Plan 01-03 quantified the mechanism structurally (data/audit/structural-audit.csv,
data/audit/structure-year-diff.csv); Plan 01-05 implemented the replicated exclusion and ran the
statewide validation (data/audit/gasb84-statewide-check.csv). This record is the ADR that closes
SRC-05.
Decision
TaxpayerScope replicates the custodial exclusion itself, applying one uniform rule to every
government and every fiscal year: the GASB-84-adjusted total for a government-year is the sum of
every fund-type column on its EDR-compiled sheet except the Custodial column, resolved by
column label (never by position). This rule is a structural no-op on any pre-FY2021 sheet, which
has no Custodial column at all — the adjusted total equals the unadjusted total for those years
automatically, with no separate branch needed.
The replicated statewide roll-up (summing the adjusted total across all 67 governments’ expenditures files, per fiscal year) is validated against EDR’s own published adjusted statewide totals as an independent confirmation — not as the target the exclusion rule is tuned to match. This choice keeps exactly one derivation path inside one series: every government-year in TaxpayerScope’s published figures is computed by the same rule, whether or not EDR happened to publish a separate per-county adjusted figure for that government and year.
Any government-year whose custodial fund share of its unadjusted total exceeds 2% is flagged.
This threshold and its flag carry forward into Phase 5’s data-quality checks (SRC-05’s disposition
in .planning/research/PITFALLS.md recommends exactly this class of check).
Rationale
Rejected alternative: use EDR’s own published adjusted figure where one exists for a given per-county file, falling back to a self-computed adjustment only where EDR does not publish one. This costs two things: it creates two derivation paths inside one series — some government- years computed by EDR’s own (opaque) adjustment logic, others by TaxpayerScope’s own rule — which means a reader cannot audit every published figure against the same stated method; and it makes TaxpayerScope’s own coverage dependent on whether EDR happened to publish an adjusted figure for a specific per-county file, a fact outside this project’s control. Replicating the exclusion uniformly avoids both costs at the price of doing the (mechanically simple, label-resolved) work ourselves.
Magnitude, one government-year: Orange County FY2021 (fl-edr-orange-expenditures, sheet
2021). Custodial fund total: $3,079,379,269.00. Unadjusted (all-fund-types) total:
$6,742,981,508.00. Custodial share: 45.67% — matching PITFALLS.md’s own independently-
sourced citation that Custodial funds were “45.7% of Orange County’s FY2020-21 published total.”
This is not a marginal adjustment: for the government this project reconciles most deeply,
custodial funds are nearly half of the sheet’s own unadjusted grand total.
The 2% flagging threshold is not a rare edge case. Rolled up across all 67 governments’
expenditures files, the count of government-years whose custodial share exceeds 2% grows from
0 of 67 in every fiscal year FY2015-2020 (no Custodial column exists before FY2021, so the
no-op rule above produces a custodial share of exactly 0.0 for every government in those years) to
35 of 67 in FY2021, 53 of 67 in FY2022, 66 of 67 in FY2023, and 66 of 67 in FY2024
(data/audit/gasb84-statewide-check.csv). Among flagged governments, the custodial share itself
is typically large, not marginally over the threshold: the median flagged government’s custodial
share was 45.63% in FY2021, 42.72% in FY2022, 43.35% in FY2023, and 41.33% in
FY2024 — figures computed directly from the flagged_governments column of the committed
dataset. By FY2023-2024, custodial fund activity exceeding 2% of a government’s reported total is
the norm across Florida counties, not the exception the 2% threshold’s name might suggest.
Source Evidence
The measured mechanism
Verified directly against the archived Orange County workbook (Plan 01-01’s tracer, extended by
Plan 01-03’s structural audit across all 67 governments): the Custodial column first appears in
the FY2021 sheet of every one of the 67 governments’ expenditures files
(data/audit/structural-audit.csv: fiduciary_column_count is 3 for all 134 government-series
rows in every fiscal year FY2015-2020, and 5 for the great majority of government-series rows
FY2021 onward — see data/audit/structure-year-diff.csv’s boundary analysis below for the
handful of exceptions). The sheet’s own “Total - All Account Codes” row is a plain SUM() across
every fund-type column including Custodial where present — verified this task by summing
Orange County’s own FY2021 fund-type columns and confirming the total matches the sheet’s own
printed grand total to the dollar. The fiduciary column set grows from 3 columns
(Pension, Trust, Component Units) pre-FY2021 to 5 columns (Custodial, Pension,
Trust, Private Purpose, Component Units) FY2021 onward, in the exact same revision that
relabels the total column’s header text (Account Total to Total Account).
The statewide validation table
Every row below is quoted verbatim from data/audit/gasb84-statewide-check.csv (committed).
replicated_adjusted_total is TaxpayerScope’s own roll-up (all 67 governments’ EDR-compiled
expenditures files, Custodial column excluded by label). edr_adjusted_total and
edr_unadjusted_total are EDR’s own published figures for the same fiscal year, read directly
from EDR’s per-fiscal-year statewide reference workbook (STATEWIDE_ADJUSTED_SOURCE_SPECS,
archived and hashed in data/sources/fl-edr-sources.json).
| FY | Replicated adjusted total | EDR adjusted total | EDR unadjusted total | Absolute difference | % difference |
|---|---|---|---|---|---|
| 2015 | $42,127,497,889.00 | $37,648,543,604.00 | $37,648,543,604.00 | $4,478,954,285.00 | 11.8968% |
| 2016 | $43,633,688,811.00 | $38,724,729,622.00 | $38,724,729,622.00 | $4,908,959,189.00 | 12.6765% |
| 2017 | $44,200,055,654.00 | $39,394,697,744.00 | $39,394,697,744.00 | $4,805,357,910.00 | 12.1980% |
| 2018 | $46,480,437,956.00 | $41,763,404,749.00 | $41,763,404,749.00 | $4,717,033,207.00 | 11.2947% |
| 2019 | $49,436,238,577.00 | $44,232,993,521.00 | $44,232,993,521.00 | $5,203,245,056.00 | 11.7633% |
| 2020 | $53,878,013,512.00 | $48,804,501,280.00 | $48,804,501,280.00 | $5,073,512,232.00 | 10.3956% |
| 2021 | $57,858,342,137.00 | $52,561,767,445.00 | $77,064,319,712.00 | $5,296,574,692.00 | 10.0769% |
| 2022 | $60,098,805,741.94 | $53,903,070,528.94 | $87,376,110,146.94 | $6,195,735,213.00 | 11.4942% |
| 2023 | $64,967,198,038.00 | $58,547,139,192.00 | $110,894,156,899.00 | $6,420,058,846.00 | 10.9656% |
| 2024 | $70,416,400,775.00 | $63,699,846,975.00 | $113,108,226,108.00 | $6,716,553,800.00 | 10.5441% |
The two do not agree exactly — the shape of the disagreement is understood and its cause is named, not left open. In every one of the ten fiscal years, TaxpayerScope’s replicated 67-government roll-up exceeds EDR’s own published adjusted statewide total by an amount that equals Duval/Jacksonville’s own adjusted total for that year, to the dollar — for example, in FY2021 the difference is $5,296,574,692.00, and Duval/Jacksonville’s own adjusted total for FY2021, computed by the identical exclusion rule, is $5,296,574,692.00. This is because EDR’s own per- fiscal-year statewide reference workbook’s “Statewide” column is a sum of 66 county columns only — confirmed this task by independently summing all 66 of that workbook’s county columns and finding they equal its own “Statewide” total exactly. Duval/Jacksonville’s data is structurally absent from that specific reference workbook (it is filed on EDR’s separate municipal fiscal page, per ADR-0004), so EDR’s own “statewide” figure never included it in the first place. This is the same finding, from the same underlying cause, that ADR-0004 records for the identifier scheme: EDR’s own published statewide reference simply omits Duval/Jacksonville by construction. TaxpayerScope’s roll-up correctly includes all 67 governments this project tracks, which is why it is larger by exactly Duval’s own contribution every year. This exclusion rule was not adjusted to close this gap — the gap is explained by a difference in scope (67 governments versus 66), not by a flaw in the replicated custodial-exclusion mechanism, and every dollar of the difference is accounted for by a named, independently-verified cause.
For fiscal years FY2015-2020, the edr_adjusted_total and edr_unadjusted_total columns are
identical — consistent with the GASB 84 boundary at FY2020-21: no Custodial column, no
adjustment, no-op.
Methodology Version
This decision takes effect in methodology_version: 1. It has no prior version to supersede —
this is the first record of the GASB 84 custodial-fund series decision.
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.