# Methods and source reconciliation

## Scope and coverage

The financial panel has 55 city-years, FY2015–2025, for Austin, Dallas, Fort Worth, Houston and San Antonio. It contains 561 observed metric records and 11 explicit missing Fort Worth sales-tax records, 572 rows in total. Eight core measures are available for every city-year. Additional accrual tax measures preserve their own scope. The pension companion has 120 plan-year observations for 12 plans, covering city reporting years FY2016–2025. It does not add a FY2015 pension baseline. BRB principal is a FY2025 cross-section, not a completed debt-stock history.

Only public records were used. The ten ACFR PDFs are original downloads from city websites or their linked document host. FY2025 statistical tables supply FY2016–2025; FY2024 supplies FY2015. Other retained originals are BLS CPI responses, Census Texas population files, BRB issuer profiles and the Comptroller city transparency guide. `sources/manifest.json` records URLs, resolved URLs, UTC retrieval times, bytes, SHA-256, document dates where identified, fiscal periods, content types and HTTP modification dates. Missing publication dates remain null with an explanatory status. HTTP modification dates are not treated as publication dates.

The statistics and required supplementary pension schedules are unaudited. Inclusion within an audited ACFR does not extend the audit opinion to those schedules. Thirty FY2025 financial cells were separately checked against rendered basic statements; the historical series have not all been traced back to their contemporaneous audited basic statements.

## Fiscal years and reporting perimeters

Houston's year ends June 30. The others end September 30. All rows retain the exact end date; the analysis compares fiscal-year labels without claiming identical economic windows. It does not convert budgets or projections into actual results.

Governmental-fund flows use modified accrual and current financial resources. Governmental and business-type net position use full accrual and economic resources. Transfers and borrowing in other financing sources are excluded from the governmental-fund revenue denominator. A governmental-fund expenditure-minus-revenue gap includes capital spending and financing arrangements; it is not automatically an operating deficit.

Primary-government activities and discretely presented component units remain separate. Austin includes electric and water operations in primary-government business activities. San Antonio reports CPS Energy and SAWS as component units. Dallas sanitation moved to enterprise funds in FY2016; stormwater moved in FY2024. Its balance history restates FY2023 for stormwater while its governmental-fund flow table explicitly leaves FY2023 un-restated. Fort Worth added MedStar EMS operations on July 1, 2025; the transfer is disclosed in Note A.15, physical PDF pages 87–88. Published values are retained rather than manufacturing a constant-boundary series.

GASB 75 introduced a break in OPEB recognition around FY2018. Lease recognition under GASB 87 and subscription liabilities under GASB 96 also change the reported debt perimeter; implementation timing differs across cities. FY2025 compensated-absence changes and other restatements can affect opening net position. Those effects must be separated from operating changes before an unrestricted-net-position trend is described as financial improvement.

## Units and inflation

City table dollars are converted from thousands to dollars exactly once. Percentages stay in percentage points on a 0–100 scale. BRB values are already dollars; Census observations are persons.

BLS series CUUR0000SA0 is CPI-U, all items, US city average, not seasonally adjusted, 1982–84 = 100. Flows use the mean of all twelve monthly CPI observations within each city's fiscal year. The common flow price base is the mean from October 2024 through September 2025. Thus Houston's FY2025 flow is also adjusted to that common base rather than receiving an automatic factor of one. Stocks use the fiscal-ending month's CPI and a common September 2025 base. Ratios are not deflated. No missing CPI month within an included fiscal window is imputed; the script requires all twelve.

CPI is a general purchasing-power proxy, not a municipal wage, construction or service-cost index. Interest costs, capital needs and local input costs may move differently.

## Population

Census incorporated-place rows are selected by summary level 162, Texas state FIPS 48 and exact place codes: Austin 05000, Dallas 19000, Fort Worth 27000, Houston 35000 and San Antonio 65000. The join is one row per city-year and is validated as many-to-one from financial observations.

FY2015–2019 use Vintage 2019 July 1 estimates. FY2020–2025 use Vintage 2025. The 2019–2020 link crosses census benchmarks and estimate vintages, so it is not a pure annual migration estimate. The chart breaks the line there. The July 1 estimate is a resident-population proxy for the fiscal-year label, including Houston's June 30 year. Service populations, commuters and utility customers can differ materially from residents.

## Historical vintage comparison

All 459 comparable FY2016–2024 metric-year values were compared between the two ACFR vintages. `vintage_differences.csv` lists the 24 differences. The panel consistently selects the FY2025 schedule vintage for that overlap; this is a published-vintage choice, not proof that every series is retrospectively harmonized.

Differences include Dallas and Fort Worth unrestricted net-position figures, three Houston unrestricted net-position figures, and historical debt ratios. Houston's net-position differences equal the firefighters' reported net pension assets in the corresponding years, which suggests a presentation relationship; that is an inference, not a completed accounting explanation. The precise correction/presentation bridge for each changed net-position cell remains open.

Fort Worth's FY2016–2023 published debt-service percentages fail a direct recalculation using the principal, interest/issuance costs and noncapital-spending amounts in the same source tables. The differences are preserved in `fort_worth_debt_service_recalculation.csv` and the eight affected panel rows are marked `observed_source_ratio_conflict`; they are not silently corrected. FY2015, FY2024 and FY2025 agree at the published precision. Historical affected percentages are excluded from trend conclusions. The FY2025 cross-section uses the printed ratios, with their city-specific definitions stated.

Houston defines noncapital expenditures using capitalized portions of capital outlay and functional expenditures. Other cities' definitions and treatment of issuance costs, leases and subscriptions vary. The published percentages should not be treated as a uniform debt-service-to-recurring-revenue ratio.

## Pensions

The companion retains NPL, total pension liability and fiduciary net position as separate source observations, checks TPL minus fiduciary net position against NPL within rounding, and calculates funding ratios from those inputs. Negative NPL values are assets, not missing data. Measurement dates remain explicit: generally December of the preceding calendar year for Austin, Dallas and the selected San Antonio plans; preceding September for Fort Worth; current June for Houston.

Pension NPL is an actuarial/accounting stock, not the next year's cash contribution. Discount rates, benefit terms, asset returns, recognition rules and measurement timing can all move it. No common-discount-rate restatement or stress test is claimed.

San Antonio utility plans are outside the selected plan set. Houston municipal RSI retains its reported plan population; a separate allocation is necessary to equate it with the primary-government balance-sheet liability. The plan series should therefore not be indiscriminately summed into city fiscal-health scores. Dallas's contribution comparison is actual versus the actuarially determined amount in FY2025 RSI, not a legal compliance conclusion.

## BRB and Comptroller sources

BRB issuer profiles supply principal classified as tax supported or revenue supported at August 31, 2025. The sum of those two categories is checked against each total. BRB principal is not ACFR carrying value, annual debt service, total long-term liabilities or NPL. School districts, counties and overlapping governments are outside the selected city profiles. Conduit/lease coverage and issuer aliases need explicit review before a complete exposure claim. The Comptroller guide supplies reporting context, not a replacement financial metric.

## Exact remaining acquisition and reconciliation

1. Follow the Download CSV links in each retained BRB issuer profile for principal history, purpose and scheduled debt service. Issuer IDs are Austin 137, Dallas 833, Fort Worth 1199, Houston 1787 and San Antonio 3226. The Dallas history link is `https://data.brb.texas.gov/charts/local/city/833/local_issuer_debtoutstanding_longitudinal_go_rev.csv`; follow each source's actual link rather than assuming another alias. Reconcile issuer/component perimeters and each city's ACFR debt note, including premiums, refunds, leases and subscriptions, at aligned dates. A full debt-stock panel is not yet delivered.
2. For Fort Worth, acquire the contemporaneous FY2015–FY2024 governmental-fund revenue statements and tax notes; combine sales-tax rows only after reconciling Crime Control and Prevention District and other special sales taxes. FY2025 basic statement physical page 53 already separately identifies $366.657m of governmental-fund sales taxes at modified-accrual scope. That cannot be silently inserted into the accrual sales-tax series.
3. Use Dallas FY2016 and FY2024 reclassification notes and enterprise statements to backcast sanitation and stormwater flows. Preserve reported and harmonized series side by side.
4. Trace the four cities' FY2024–2025 intergovernmental declines to the grants, ARPA and nonmajor-fund schedules, then classify recurring and one-time recognition. An aggregate decline is not a verified program sunset.
5. Resolve each net-position vintage difference against correction/restatement notes and contemporaneous basic statements. Ask the city for clarification of the historical Fort Worth ratio row only if an authorized later workflow permits outreach; no outreach occurred here.
6. To add a FY2015 pension baseline, extract the earliest GASB 68 pension schedules from the retained FY2024 reports and reconcile initial-adoption/measurement dates. The current pension companion already provides a consistent ten-reporting-year window.

## Verification and delivery

Scripts use local reviewed files and deterministic joins. Hash checks establish source identity; row-count and uniqueness checks establish coverage; independent basic-statement and pension identities establish specific calculation checks. None of these is an accuracy score for the entire subject. Figures were inspected as rendered PNGs, with source labels, units, dates and relevant qualifications visible. The portable bundle contains data, scripts, selected text extracts, manifests, figures and this memo; the larger original PDFs remain in the assigned research area and can be retrieved with frozen hash verification.