# Responsible America — Middle-Class Historical Timeline Methodology Status: architecture document plus a first real, hand-verified seed dataset (added 2026-08-11). Companion to `docs/public-scorecard-methodology.md`, `docs/political-money-tracking-methodology.md`, `docs/federal-contracts-tracking-methodology.md`, `docs/ig-watchdog-tracking-methodology.md`, and `docs/promise-ledger-methodology.md` (same "architecture before data" approach, and the same real-data-civic-tools batch). Last reviewed: 2026-08-11. ## 1. What this is, and what it is not This is the data architecture for a historical timeline of middle-class economic conditions ("America Then and Now"), named in Civic Clarity's own planning corpus (`civic-clarity/MASTER_VISION_RECONCILIATION_MATRIX.md` row 9) and in this task's own source prompt. `data/schemas/middle-class-timeline-series.schema.json` defines the shape a real, verified series takes. **It is still not a populated multi-decade timeline, and still not a live tracker.** As of 2026-08-11, `middle-class-timeline.html` shows a seed dataset of 4 real series (§2 below), each a one-time, hand-verified snapshot spanning 2019 through 2024 -- a defensible recent-history core, not the full historical sweep every category could eventually support, and not refreshed on any schedule. ## 2. The seed dataset: four real series (2019-2024) Added 2026-08-11 as this tool's first conversion from architecture-only to real, source-backed content. Selection criteria and the specific data-reconciliation issues found during research are documented in full in `research/tool-middle-class-timeline-seed/research-plan.md`. In short: - **Real median household income** (`income_and_wages`): 6 points, 2019-2024, all expressed in a single consistent 2024-dollar basis (FRED's re-based MEHOINUSA672N series) rather than mixing each year's own-year dollars from Census's individual press releases -- a real reconciliation issue this dataset documents rather than glosses over. - **Unemployment rate** (`employment_and_labor_force`): 6 points, 2019-2024, U-3 seasonally adjusted annual average, BLS. - **Labor force participation rate** (`employment_and_labor_force`): 5 points, 2019-2023 only -- a 2024 annual average was not confirmed with adequate specificity via search this pass, so it is left out rather than approximated from a single monthly reading. - **CPI-U annual inflation rate** (`inflation`): 6 points, 2019-2024, year-over-year percent change, BLS. **Federal debt held by the public as a percent of GDP was deliberately NOT populated this pass**, despite its source record (`src-tool-mct-cbo-federal-debt-held-by-public`) already existing: multiple WebSearch passes for FY2019-2024 actual figures returned inconsistent numbers depending on which CBO report vintage was scraped (projected vs. later-revised-actual figures for the same fiscal year). Rather than publish one arbitrarily chosen inconsistent figure, this series' `data_points` remain empty, and the reconciliation difficulty is documented directly in the research plan -- the same "don't force every planned metric" discipline this milestone applies throughout. Every data point carries the same fixed `causation_caveat: "context_not_causation"` marker every series in this schema requires; no `administration_context` entries were added to any of the 4 seed series this pass, deliberately, to avoid populating potentially error-prone party-control-of-Congress claims without dedicated verification -- a future increment should add these with their own research pass, not repurpose this one's citations for a different claim. **Verified seed dataset — last verified 2026-08-11.** This is a demonstration dataset (4 of 12 categories, a 2019-2024 window), not a comprehensive multi-decade timeline. ## 3. The 12 measure categories Grouped from the task's own list of measures: | Category | Task's named measure(s) | Authoritative producer | |---|---|---| | `income_and_wages` | real median household income; real wages | U.S. Census Bureau; Bureau of Labor Statistics | | `employment_and_labor_force` | employment; labor-force participation; unemployment | Bureau of Labor Statistics | | `inflation` | inflation | Bureau of Labor Statistics (CPI) | | `interest_rates` | interest rates | Federal Reserve Board | | `housing_costs` | housing affordability | Census Bureau; Federal Housing Finance Agency | | `healthcare_costs` | healthcare costs | Bureau of Labor Statistics; CMS | | `education_costs` | education costs | National Center for Education Statistics | | `childcare_costs` | childcare, where reliable series exist | Department of Labor; Census | | `retirement_security` | retirement security | Social Security Administration; Federal Reserve (SCF) | | `productivity` | productivity | Bureau of Labor Statistics | | `wealth_and_distribution` | wealth/distribution, where methodologically supportable | Federal Reserve Board (Distributional Financial Accounts) | | `federal_debt_and_fiscal_context` | federal debt/fiscal context | Congressional Budget Office; Treasury | Five categories (`income_and_wages`, `employment_and_labor_force`, `inflation`, `wealth_and_distribution`, `federal_debt_and_fiscal_context`) have a real, verified source record in this repository; three of those five (`income_and_wages`, `employment_and_labor_force`, `inflation`) also have real, populated `data_points` as of 2026-08-11 -- see §2. The remaining categories are named here as the intended taxonomy for future research passes, not populated with placeholder sources. ## 4. "Preserve administration/Congress context without implying simple presidential causation" This is the same structural discipline the Promise Ledger's `authority_map` enforces, applied to economic time series instead of campaign promises. Two schema-level mechanisms enforce it: 1. **`causation_caveat`** is a required, fixed-value field on every series (`"context_not_causation"`) — a marker that cannot be silently omitted, mirroring the Promise Ledger's `legal_review_status` gate pattern. 2. **`administration_context`** is deliberately separated from `data_points`: it records which administration and which party controlled Congress during a given period, for orientation only. It is never rendered adjacent to editorializing language that implies the officeholder caused the indicator's movement — a national economic indicator is shaped by monetary policy, global conditions, prior law, demographic change, and dozens of other factors no single administration controls, in addition to whatever influence that administration's own policy choices had. ## 5. Why a real measurement break matters: the 2013 CPS ASEC example The clearest concrete illustration of why `methodology_notes` exists: Census's own working paper on the 2013 CPS ASEC income-question redesign documents a measured ~3.17 percentage point increase in reported median household income for 2013 caused by the redesigned questions themselves — not by a change in the underlying economy. A timeline that showed a jump across 2012–2013 without this note would risk being read as if some policy or event caused a sudden income gain, when the actual cause was how the question was asked. See `src-tool-mct-census-cps-2013-redesign-break`. ## 6. Wealth and distribution: "where methodologically supportable" The task's own qualifier on this measure is deliberate — wealth distribution is harder to measure reliably than income, because it depends on survey-based estimates of assets and debts that are unevenly held and unevenly reported. The Federal Reserve's Distributional Financial Accounts (`src-tool-mct-fed-distributional-financial-accounts`) are selected specifically because they are a real, published, peer-reviewable methodology for this exact problem: integrating the triennial Survey of Consumer Finances with the quarterly Financial Accounts of the United States to produce a defensible quarterly wealth-distribution series back to 1989. Any future `wealth_and_distribution` series record in this repository should cite this methodology directly rather than a secondary restatement of it. ## 7. What would need to be true before this becomes a live, comprehensive timeline A real data-ingestion pass per remaining category, each data point individually sourced and dated (not a single bulk import); a defined refresh cadence matching each source's own publication schedule (e.g. Census's annual income release, BLS's monthly CPS release); extending each of §2's 4 series further back in time and to the present; and an editorial review step confirming every future `administration_context` entry is written as context, not causation, before publication — consistent with this platform's existing review-gate conventions for other sensitive content categories.