A 36-year run of Federal Reserve wealth data, projected forward by three independent forecast models. The question: in what year does the median Black household catch the median white household?
At current rates, parity does not arrive.
The series moves away from parity, not toward it.
★ Per-household net worth, real 2022 dollars · log scale ★
Forecast model
Y axis
★ Distance to parity ★
Parity · 0.95
0.00 — Total inequality1.00 — Wealth equality
★ Convergence year — when Black/White ratio reaches 0.95 ★
★ Wealth gap by decade · per-household, real 2022 dollars ★
★ Methodology · transparency before persuasion ★
How this was made
The question
At the current rate of change, in what year does the median Black household catch the median white household? The plate's argument is not in the exactness of the answer — it is in the fact that three independent forecast models, run on the federal government's own data, all return the same verdict: it does not happen.
Code path: plates/01-parity-clock/01_clean.py handles the DFA zip extract, the CPI deflation, and the parity ratio computation.
The forecast
Three independent extrapolations, all on the parity ratio (Black per-household wealth ÷ white per-household wealth):
Geometric — compound rate of change from 1989Q3 to 2025Q4, projected forward.
Linear — ordinary-least-squares regression on the full series, projected forward.
ARIMA(1,1,1) — fit via statsmodels.tsa.arima, 2,000 forward steps.
Each forecast is checked against the parity threshold (ratio ≥ 0.95) over a 500-year horizon. None converge.
A fourth model — TimesFM 2.5 (Google Research zero-shot foundation model with quantile output) — is in scope per the series spec but currently blocked on a Python 3.14 dependency conflict (jaxlib). It will be added in a follow-up build with no change to the headline finding.
What this is not
This is not a prediction. It is an extrapolation that assumes the conditions of the last 36 years continue. The plate's purpose is to indict the assumption that the status quo arrives at parity on its own. If policy changes — if reparations, baby bonds, public-option banking, or an inheritance-tax overhaul reshape the underlying mechanics — the trajectory shifts. The forecast does not predict the future. It describes a present that is not on a path to repair itself.
The license
MIT-licensed. Forkable. Built in the lineage of Anthony Starks's #DuBoisChallenge toolchain and the W.E.B. Du Bois's Data Portraits volume edited by Whitney Battle-Baptiste & Britt Rusert (Princeton Architectural Press, 2018). Repository ships open at Plate 1 release.