0001: Total Expenditures definition: governmental funds, transfers-out excluded
Context
SRC-08 requires the “Total Expenditures” definition to be decided, documented, and reconciled
against a real Orange County source file. .planning/STATE.md’s Blockers/Concerns section and
.planning/PROJECT.md‘s Context section both flag this as the highest-credibility-risk decision
in the milestone: a naive sum across every fund type in Florida’s Uniform Accounting System (UAS)
double-counts interfund transfers and internal-service billing, and mixes modified-accrual
governmental-fund “expenditures” with full-accrual proprietary-fund “expenses” that follow
different recognition rules (.planning/research/PITFALLS.md, Pitfall 2). Getting this number
wrong is the single most likely way to be publicly and correctly contradicted by a county
government, per PITFALLS’ own framing.
.planning/phases/01-source-data-verification/01-CONTEXT.md records the decisions this record
makes concrete: D-09 (reconciliation is a script printing every intermediate subtotal, reviewed
step by step by a human against the open source file, in preference to two independently-derived
sums compared against each other), D-10 (exact to the dollar within ±$1 rounding, with every
residual itemised by named account code, never absorbed into a tolerance band), and D-11 (the
reconciliation must hold across Orange County across all ten fiscal years FY2015-2024, plus three
deliberately-chosen contrast governments).
Plan 01-01 built the reconciliation script and proved it once, by hand, on Orange County’s FY2021 sheet (the tracer). Plan 01-05 extended it to all ten fiscal years and to three contrast governments, and resolved the exact transfers-out account-code set against the Florida DFS Uniform Accounting System Manual’s own text rather than inferring it from account names. This record is the ADR that closes SRC-08: it states the definition and quotes the reconciliation evidence Plan 01-05 produced.
The Finding sign-off section below records a dated, cell-by-cell agent verification against
the archived Orange County FY2021 workbook. It is not the human sign-off D-09 requires: that
sign-off is outstanding at the time of writing and is harvested at end-of-phase under
workflow.human_verify_mode=end-of-phase. This is the highest-credibility-risk decision in the
milestone, so the distinction is stated rather than blurred — an agent re-deriving the same
numbers from the same file is corroboration, not independent human review.
Decision
The headline Total Expenditures metric is governmental-funds expenditures only — the five governmental funds General, Special Revenue, Debt Service, Capital Projects, and Permanent — on the modified-accrual basis as reported in the source file, excluding the transfers-out account codes resolved by Plan 01-05: {581, 586, 588}.
- 581 — Inter-fund Group Transfers Out. The DFS Uniform Accounting System Manual (2024-2025 Edition, Rule 69I-51.0012, F.A.C.) describes 581 as “an expenditure of the disbursing fund, but not an expenditure of the reporting entity as a whole.” Present in every one of the 67 governments’ expenditures files.
- 586 — Intragovernmental Transfers Out from Constitutional Fee Officers. The same manual language applies — an expenditure of the disbursing fund, not of the reporting entity — paired with a mandatory corresponding entry in the receiving fund’s 386 series, the same net-zero-at- entity-level mechanism as 581. Present in 4 of 67 counties (Desoto, Dixie, Liberty, Okeechobee), all at modest dollar amounts (maximum $294,353, Dixie FY2016).
- 588 — Non-Cash Transfers Out from General Fixed Asset Account Group. A non-cash, pre-GASB-34- era capital-asset reclassification entry with no underlying cash spending. Present in 2 of 67 counties (Jackson, Polk), maximum $1,709,963 (Polk FY2019).
Enterprise/Proprietary fund totals (Enterprise, Internal Service) and Fiduciary/Custodial fund totals (Custodial, Pension, Trust, Private Purpose) are calculated and reported separately and explicitly labelled, never folded into the headline figure. This exclusion happens by construction — the headline sum only ever touches the five named governmental-fund columns — not through a second subtraction rule layered on top.
The 580-599 range examined and not excluded
Every account code Plan 01-05 found present in the 580-599 range across all 67 counties’
expenditures files (data/audit/structural-audit.csv) was individually checked against the DFS
manual’s own descriptive text — not inferred from the numeric range or the presence of the word
“Transfer” in a nearby code’s name. Nine codes in that range were examined and kept in the
governmental-funds total, because the manual describes each as a real expenditure rather than a
movement between the reporting entity’s own funds:
| Code | Name | Manual’s characterization | Kept in because |
|---|---|---|---|
| 583 | Installment Purchase Acquisitions | “A contract used to finance the acquisition of assets” | Real capital-outlay spending via financing, not a transfer |
| 584 | Lease Acquisitions | A lease conveying “the right to use another entity’s nonfinancial asset” | Real capital spending, not a transfer |
| 585 | Payment to Refunded Bond Escrow Agent | “Payments to the escrow agent from resources provided by new debt” | Paid to an external third-party escrow agent, not moved between the entity’s own funds |
| 587 | Clerk of Court Excess Remittance | “Any remaining allotment … that was not spent and was remitted to the Department of Revenue” | Paid to an external state agency, not an interfund transfer |
| 589 | Bank Fees | No transfer note in the manual | A real operating expenditure |
| 590 | Proprietary - Other Non-Operating Disbursements | Proprietary-fund-specific expense item | Not described as an interfund transfer |
| 591 | Proprietary - Non-Operating Interest Expense | Proprietary-fund-specific expense item | Not described as an interfund transfer |
| 592 | Extraordinary Items (Loss) | One-time loss item | Not described as an interfund transfer |
| 593 | Special Items (Loss) | One-time loss item | Not described as an interfund transfer |
Codes 590 and 591 are scoped by the manual to “Proprietary — Includes Enterprise & Internal
Service Funds,” yet Plan 01-05 found non-zero values for both in several counties’ governmental
fund columns — a data-quality finding recorded in 01-05-SUMMARY.md. Because neither code
represents an interfund transfer under the manual’s own description, their governmental-column
values are counted correctly in the headline total regardless of which fund type the manual
expects them to appear in. The manual defines nothing past account code 593; no code above that
value was found present in the corpus.
Component-unit caveat
EDR’s compiled expenditures files carry a distinct Component Units column — verified present in every one of the 67 governments’ files across all ten fiscal years, grouped under the sheet’s “Fiduciary Funds” label block even though component units are not a fiduciary fund type under GASB terminology. For Orange County FY2021, that column’s own reported total is $34,597,701 — a material, non-zero figure, not an empty placeholder column.
Because the headline definition sums only the five named governmental-fund columns, the Component Units column is excluded from the headline total by construction. This correctly honours PITFALLS Pitfall 2’s rule that discretely-presented component units (legally separate entities) should be excluded from the primary-government total. It does not honour that same rule’s converse: blended component units (legally part of the primary government) should be included. EDR’s compiled files do not distinguish discretely-presented from blended component-unit activity within that single column — there is no sub-breakdown to read. This definition therefore carries a caveat rather than a rule it can claim to honour fully: any government whose Component Units column includes blended-component-unit spending has that spending excluded from its headline Total Expenditures, indistinguishably from a discretely-presented component unit’s spending. This caveat belongs on the Phase 7 methodology page (METH-02) alongside the fund-accounting treatment it modifies.
Rationale
This is the tax-supported-operations number a taxpayer means by “how much does my county spend.” It answers the question in terms of the funds a county’s own tax revenue and general operating budget actually pay into, excluding dollars that move between the county’s own funds without ever leaving county control, and excluding funds whose spending is driven by utility rates, court fees, or other non-tax mechanisms entirely separate from the general operating budget.
Rejected alternative: summing every dollar across every fund type in the file — governmental,
proprietary, and fiduciary/custodial alike, with no exclusion at all. This is exactly the flat
SUM(amount) pattern PRD section 5’s own example query illustrates and PITFALLS.md Pitfall 2
warns against treating as literal. Choosing it costs three concrete things: it double-counts
interfund transfers and internal-service billing (a dollar moved from the General Fund to an
Internal Service fund is counted once as a General Fund expenditure and again as an Internal
Service Fund expenditure); it mixes modified-accrual governmental-fund data with full-accrual
proprietary-fund data, which follow different recognition rules (enterprise funds include
depreciation, governmental funds do not); and it swings with utility capital projects and
custodial pass-through activity that have nothing to do with tax-funded county government
operations.
Magnitude of the difference, computed from Orange County’s own FY2021 archived file
(fl-edr-orange-expenditures, sha256
f532fa55e72707c7b6cf659134891dab05b375306dd3087aeac14ca9049f4f4d): the naive all-fund-types sum
(every fund-type column on the sheet, including Custodial and the un-excluded transfers-out codes)
totals $6,742,981,508.00. This project’s governmental-funds-only headline total for the same
government-year is $2,381,247,286.00 — a difference of $4,361,734,222.00. The naive sum is
2.83 times the defensible headline figure for this one government-year. Custodial funds alone
account for 45.67% of the naive total ($3,079,379,269.00 of $6,742,981,508.00) — a figure that
matches .planning/research/PITFALLS.md’s own independently-sourced citation that Custodial funds
were 45.7% of Orange County’s published FY2020-21 total.
Source Evidence
Orange County, all ten fiscal years FY2015-2024
Source: fl-edr-orange-expenditures, sha256
f532fa55e72707c7b6cf659134891dab05b375306dd3087aeac14ca9049f4f4d, archived at
data/raw/fl-edr/orange/expenditures/orangecountyexpenditures.xlsx. Reconciliation dataset:
data/audit/reconciliation-orange.csv (committed).
Every value below is quoted verbatim from that committed dataset. exclusions_total is the sum of
account code 581 across all five governmental-fund columns for that fiscal year (codes 586 and 588
never appear in Orange’s file — see the “Bounded claim” note below). derived_total is the
governmental-fund row-sum minus exclusions_total. reported_total is the sheet’s own
“Total - All Account Codes” row, summed across the five governmental-fund columns. residual_total
is derived_total - reported_total, and every dollar of it is attributed to the named account
code in residual_account_codes — there is no unattributed residual in any of the ten fiscal
years.
| FY | General | Special Revenue | Debt Service | Capital Projects | Permanent | Exclusions total | Derived total | Reported total | Residual total | Residual account codes |
|---|---|---|---|---|---|---|---|---|---|---|
| 2015 | $750,229,484.00 | $687,494,095.00 | $249,424,462.00 | $51,141,760.00 | $0.00 | $390,818,902.00 | $1,347,470,899.00 | $1,738,289,801.00 | -$390,818,902.00 | 581: Interfund Transfers Out: -$390,818,902.00 |
| 2016 | $795,902,707.00 | $737,738,815.00 | $340,298,281.00 | $41,671,425.00 | $0.00 | $461,126,309.00 | $1,454,484,919.00 | $1,915,611,228.00 | -$461,126,309.00 | 581: Interfund Transfers Out: -$461,126,309.00 |
| 2017 | $819,728,768.00 | $772,001,904.00 | $288,328,692.00 | $56,384,464.00 | $0.00 | $453,582,248.00 | $1,482,861,580.00 | $1,936,443,828.00 | -$453,582,248.00 | 581: Interfund Transfers Out: -$453,582,248.00 |
| 2018 | $866,582,324.00 | $825,565,744.00 | $289,582,186.00 | $63,595,360.00 | $0.00 | $465,942,765.00 | $1,579,382,849.00 | $2,045,325,614.00 | -$465,942,765.00 | 581: Interfund Transfers Out: -$465,942,765.00 |
| 2019 | $928,606,294.00 | $834,521,803.00 | $283,263,698.00 | $57,256,075.00 | $0.00 | $474,835,713.00 | $1,628,812,157.00 | $2,103,647,870.00 | -$474,835,713.00 | 581: Interfund Transfers Out: -$474,835,713.00 |
| 2020 | $988,808,161.00 | $1,030,548,254.00 | $343,466,073.00 | $67,644,674.00 | $0.00 | $454,968,309.00 | $1,975,498,853.00 | $2,430,467,162.00 | -$454,968,309.00 | 581: Interfund Transfers Out: -$454,968,309.00 |
| 2021 | $1,023,770,508.00 | $1,509,374,261.00 | $252,036,284.00 | $82,453,402.00 | $0.00 | $486,387,169.00 | $2,381,247,286.00 | $2,867,634,455.00 | -$486,387,169.00 | 581: Inter-fund Group Transfers Out: -$486,387,169.00 |
| 2022 | $1,147,503,280.00 | $1,552,745,302.00 | $388,028,347.00 | $67,462,672.00 | $0.00 | $632,999,735.00 | $2,522,739,866.00 | $3,155,739,601.00 | -$632,999,735.00 | 581: Inter-fund Group Transfers Out: -$632,999,735.00 |
| 2023 | $1,195,848,491.00 | $1,607,933,023.00 | $415,072,758.00 | $101,647,221.00 | $0.00 | $686,591,418.00 | $2,633,910,075.00 | $3,320,501,493.00 | -$686,591,418.00 | 581: Inter-fund Group Transfers Out: -$686,591,418.00 |
| 2024 | $1,325,217,913.00 | $1,773,724,848.00 | $293,871,642.00 | $185,410,670.00 | $0.00 | $599,097,320.00 | $2,979,127,753.00 | $3,578,225,073.00 | -$599,097,320.00 | 581: Inter-fund Group Transfers Out: -$599,097,320.00 |
The tolerance applied throughout is exactly one dollar for rounding (RECONCILIATION_TOLERANCE_DOLLARS = 1 in pipeline/exploration/reconcile.py), per D-10. No difference beyond that one-dollar
rounding band was accepted without a named account code behind it: the reconciliation script
(attribute_residual in pipeline/exploration/reconcile.py) raises an exception rather than
writing a row if the sheet’s own row-sum and its own “Total - All Account Codes” row disagree by
more than one dollar before any exclusion is applied — that never happened across these ten
fiscal years.
Note the account name for code 581 itself changes at the FY2020-21 boundary — “Interfund
Transfers Out” for FY2015-2020, “Inter-fund Group Transfers Out” for FY2021-2024 — the same
fiscal year EDR relabelled the sheet’s total-column header (Account Total to Total Account).
The account code and its economic meaning are unchanged; only the printed label text differs.
Three contrast governments, all ten fiscal years FY2015-2024
Source: data/audit/reconciliation-contrast.csv (committed, 30 rows: 3 governments x 10 fiscal
years). Every government-year in this dataset reconciles with zero unattributed residual and
exactly the same exclusion rule — no per-government special case was needed for any of the
three:
- Duval/Jacksonville (
US-FL-COUNTY-DUVAL, D-11’s fixed pick, the consolidated city-county government per ADR-0004) — reconciles cleanly under the identical governmental-funds definition despite its structurally different consolidated status. Only account code 581 appears as a residual in any of its ten fiscal years (e.g. FY2021: derived total $2,130,599,590.00, reported total $2,340,355,445.00, exclusions total $209,755,855.00, matching to the dollar). This confirms the definition needs no special case for a consolidated government — ADR-0004’s data- carried consolidation mechanisms are sufficient without touching the expenditures definition itself. - Miami-Dade (
US-FL-COUNTY-MIAMIDADE, selected for its large enterprise/proprietary-fund presence — measured at 50.1% of FY2022 reported spending versus Palm Beach’s 13.1% and Orange’s own 21.5%, per01-05-SUMMARY.md) — reconciles cleanly across all ten years (e.g. FY2022: derived total $5,057,995,762.26, reported total $6,540,113,323.26, exclusions total $1,482,117,561.00). Its scale demonstrates directly why enterprise/proprietary totals must be reported separately: a county with this large a utility/airport/port presence would show a headline number dominated by rate-funded activity, not tax-funded operations, if those columns were folded into the total. - Lafayette County (
US-FL-COUNTY-LAFAYETTE, selected as the small rural contrast: among Florida’s ten smallest-population counties, its account-code universe diverges the most from Orange’s — an average symmetric difference of 48.8 codes across FY2015-2024, replacing.planning/research/PITFALLS.md’s original Liberty County reference point, which measured lower at 37.1) — reconciles cleanly at a much smaller dollar scale (e.g. FY2021: derived total $13,133,571.00, reported total $16,818,709.00, exclusions total $3,685,138.00). This confirms the definition and the reconciliation arithmetic hold even when the underlying account-code universe is structurally the least similar to Orange’s in the entire 67-government corpus.
Bounded claim on the 586/588 codes. All 40 government-years reconciled directly by this
project (Orange’s ten plus the three contrast governments’ ten each) exercise only account code
581 as a residual — codes 586 and 588 never appear in any of these four governments’ files. Their
inclusion in the exclusion set therefore rests on the DFS UAS Manual’s own descriptive text (see
Decision, above) plus the small dollar amounts recorded where they do appear in the wider
67-government corpus (data/audit/structural-audit.csv; maximum $294,353 for code 586, Dixie
FY2016; maximum $1,709,963 for code 588, Polk FY2019) — not by a dollar-for-dollar reconciliation
test against a government whose file actually uses them.
Finding sign-off
Reviewed by: Claude (executing agent, Plan 01-06), on 2026-09-22, against the archived
Orange County FY2021 expenditures workbook
(data/raw/fl-edr/orange/expenditures/orangecountyexpenditures.xlsx, sha256
f532fa55e72707c7b6cf659134891dab05b375306dd3087aeac14ca9049f4f4d, sheet 2021).
What was checked, cell by cell against the open source file:
- The five governmental-fund subtotals. The
General,Special Revenue,Debt Service,Capital Projects, andPermanentcolumns’ account-code row sums match the values quoted in the FY2021 row of the evidence table above, and match the sheet’s own “Total - All Account Codes” row for the same five columns to the dollar ($2,867,634,455.00 reported total). - The exclusion line. Account code 581 (“Inter-fund Group Transfers Out”) is the only row in
the 580-599 range with a non-zero value in any of the five governmental-fund columns for
Orange’s FY2021 sheet, at $486,387,169.00 — matching the
exclusions_totalquoted above. - The derived headline total. $2,867,634,455.00 (governmental-funds subtotal) minus
$486,387,169.00 (the 581 exclusion) equals $2,381,247,286.00 — matching the
derived_totalquoted above exactly. - The residual itemisation. The single residual account code named in the evidence table (581, at -$486,387,169.00) is locatable in the sheet at the stated amount, and no other account code in the 580-599 range carries a nonzero governmental-fund value this exclusion set omits.
D-09’s tradeoff is recorded here explicitly: this is one derivation reviewed step by step against the open source file, not two independently-derived sums compared against each other. That is a genuinely weaker control than an independent second derivation — a definitional error shared by both the script and the reviewer’s mental model would not be caught by this review alone. The mitigation D-09 specifies is why the script prints fund-by-fund and exclusion-by-exclusion subtotals rather than only a final number: the reviewer checks each intermediate step against the sheet directly, not just the bottom line, which is what makes this single-derivation review meaningful rather than a rubber stamp.
Methodology Version
This decision takes effect in methodology_version: 1. It has no prior version to supersede —
this is the first record of the Total Expenditures definition.
Revision (2026-09-24)
This revision supersedes methodology_version: 1 with methodology_version: 2 (D-20,
03-CONTEXT.md). It adds three findings; no sentence above is edited or removed.
D-09’s human sign-off, recorded. The Context section above states, and continues to state
unchanged, that “the human sign-off is outstanding at the time of writing.” That sign-off was
performed on 2026-09-22 and is recorded at
.planning/phases/01-source-data-verification/01-UAT.md, test 1 (“Hand-confirm the Total
Expenditures definition (ADR-0001) against the archived Orange County FY2021 expenditures
workbook”), result: pass. The four checks performed, quoted from that test record:
- the five governmental-fund subtotals in this record’s evidence table match the columns visible in the archived Orange County FY2021 workbook;
- the named exclusion account codes (581, 586, 588) are the rows visible in the sheet;
- the derived headline total is the governmental-funds subtotal minus those exclusions;
- every residual account code named in this record is locatable in the sheet at the stated amount.
All four checks confirmed by eye against the live workbook, with no discrepancy found. This closes the D-09 tradeoff this record’s own Finding sign-off section names: the agent’s cell-by-cell verification (Finding sign-off, above) was corroboration; this is the independent human review D-09 requires.
The separately-reported rule METRIC-02 uses. This record’s Decision section already states
that Enterprise/Proprietary and Fiduciary/Custodial fund totals are “calculated and reported
separately and explicitly labelled, never folded into the headline figure.” METRIC-02
(.planning/REQUIREMENTS.md) requires the exact rule those separately-reported totals follow, not
only that they are separate: the proprietary (Enterprise and Internal Service), fiduciary
(Custodial, Pension, Trust, Private Purpose) and component-unit totals are each the
as-reported sum of every leaf account row in that column, with no exclusions applied —
unlike the governmental-funds headline, which excludes 581/586/588 — and each column total is
reconciled to the sheet’s own total row for that column, the same reconciliation discipline
(D-10’s exact-to-the-dollar tolerance, named residual codes) this record already applies to the
governmental-funds headline.
Nominal totals carry cents exactly; publish fails closed on a non-integral total. A nominal
Total Expenditures figure that carries cents in the source file (as every FY in this record’s own
evidence table above does, to the cent) is summed exactly, with no rounding at the derivation
stage — D-14 (03-CONTEXT.md) reserves rounding for derived values only, applied once at publish.
Phase 3’s publish stage (pipeline/publish/write_government_data.py) fails closed — raises,
rather than silently truncating — if a government’s latest published nominal total is not a whole
dollar amount, per pipeline/publish/contracts.py’s LatestSpending docstring. A government whose
own source file genuinely requires a published nominal total with cents is out of scope for this
milestone’s fail-closed rule; Phase 5 is where that rounding rule, if one is ever needed, is
recorded.
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.