Radical Imagination ↗
★ A Living Methodology Artifact · May 2026 → Dec 2026

Data Portraits
Forecasts

The Subject
Long-run public datasets of American racial inequality, projected forward at the current rate of change.
The Wager
That four plates, run on the federal government's own data, surface a verdict the public discourse refuses to name.

Du Bois's 1900 Paris plates made the present legible at the scale he could draw it. This series makes the future legible at the current rate of change — and asks whether that rate is acceptable.

The forecast is not a prediction. It is an extrapolation from 36 years of federal data, drawn in the visual grammar Du Bois set in 1900. The answer, almost always, is never, or centuries from now.

★ The forecast is the indictment ★
★ Plate I
The Parity Clock
Live · Apr 2026
Plate II
Reframe pending
Pulled · CV review
Plate III
The Decarceration Curve
Target · Jun 2026
Plate IV
The Counterfactual Engine
Target · Dec 2026
IPlate · 1 of 4 · live
After Du Bois's economic accounting, projected forward by three independent models

The Parity Clock

At current rates, the median Black household catches the median white household in the year — never. All three forecast models agree: parity does not arrive within 500 years. The Black/White wealth ratio went from 0.263 (1989Q3) to 0.220 (2025Q4). The series moves away from parity, not toward it.

Range
1989Q3 — 2025Q4 · 146 quarterly observations
Sources
Federal Reserve Distributional Financial Accounts · Survey of Consumer Finances · CPI-U deflator (BLS)
Models
Linear (OLS) · Geometric · ARIMA(1,1,1) · TimesFM 2.5 (scoped, env-blocked, follow-up build)
Status
★ Live Drafted 2026-04-13 · interactive web + static plate
Enter the plate
IIPlate · 2 of 4 · pulled
★ Pulled 2026-04-14 · Cultural Validator review in progress

Pulled for review

A plate briefly shipped at this slot on 2026-04-14 under the title Reparations ETA. The framing abstracted the HBCU / Ivy endowment gap as a proxy for reparations — a reductive frame that misstates what reparations actually demand (stolen labor, land, generational wealth, Jim Crow, redlining, incarceration). An institutional endowment comparison is not a reparations accounting.

Pulled
2026-04-14 · same day published
Reason
Reductive framing · conflated a single institutional disparity with the concept of reparations
Process failure
Cultural Validator was not invoked before public deploy. A reframed version returns only after documented CV pass.
Reframe
Target title: The Endowment Divide (or defensible alternative) · explicit methodology note that this is one institutional marker, not a reparations accounting
Read the pull notice
IIIPlate · 3 of 4 · forthcoming
After Du Bois's plate "The Negro in Prison" (American Negro Exhibit, 1900)

The Decarceration Curve

Three paces — reform, current, backlash — projected forward against the question: in what year does racial parity in incarceration arrive? Layered with 50 per-state plates, each a small mirror of the national curve.

Range
1978 — 2023 · annual + county-level overlays
Sources
BJS Prisoners Series · Vera Institute county-level incarceration trends
Deliverables
National three-paces plate · 50 state plates · interactive selector
Status
Forthcoming Target ship · Jun 2026
In development
IVPlate · 4 of 4 · thesis · Schomburg residency
The thesis piece — what was owed

The Counterfactual Engine

Train TimesFM on truncated historical windows of Black economic progress before interruption. Forecast forward to 2024 as if the interruption hadn't happened. The gap between the gold counterfactual line and the vermillion actual line is the cost — quantified in dollars. Three sub-plates, each tied to a SAM in the Constellation.

Sub-plates
Reconstruction (1865–1877) · land, literacy, wages → Douglass SAM
Tulsa Greenwood (1900–1921) · Black business formation → Wells SAM
Northern wages (1940–1968) · redlining + deindustrialization → Garvey SAM
Method
TimesFM trained on pre-interruption window · forecast continues the trajectory · gap to actual = the cost
Status
Forthcoming Target ship · December 2026 (Schomburg residency)
In development

Built in the lineage

Each plate uses the color discipline Du Bois set in 1900 — vermillion, gold, ink, cream — and the same insistence that the data be drawn plainly. The methodology page on every plate publishes the data sources, the forecast models, and the assumptions in full. No figure that cannot be cited will appear.

The toolchain is open-source, set by Anthony Starks for the #DuBoisChallenge community: dubois-data-portraits, decksh, deck. The forecast engine is TimesFM 2.5 (Google Research, zero-shot foundation model with quantile output) — supplemented with statsmodels ARIMA, OLS regression, and geometric extrapolation as cross-checks on every claim.

Whitney Battle-Baptiste & Britt Rusert's W.E.B. Du Bois's Data Portraits: Visualizing Black America (Princeton Architectural Press, 2018) is the lineage being claimed. This series is its forward-looking companion.

Forecasts ships open. MIT-licensed. Forkable. Contributable.