One hundred years of data. Analyzed. Visualized. Open.
A reporting series. Ten datasets, each spanning at least a century of American life. Each one a question the historical record can finally answer. Independent. No ads, no tracking, no paywall.
The Archive · Ten IssuesHover for the finding
About the project
Why this exists.
One Hundred Years is an independent research and visualization project. Each issue takes a single dataset — one that spans at least a century of American life — and asks what patterns become visible only at that scale.
The project is built by one person. There is no newsroom, no institution, no funding. Every dataset is public. Every methodology is published. Every line of code is open.
Press & contact
For press inquiries, interviews, or collaboration proposals:
You are free to excerpt, quote, and embed any visualization with attribution.
Methodology
How the analysis works.
Every issue uses a different analytical approach matched to its dataset. The models below are the primary engines. Each issue publishes its own methodology tab with full technical detail, known limitations, and confidence ratings.
M1Station Anomaly Engine
Per-station daily anomalies computed against a 1901–2000 baseline. Each station scored against its own historical distribution. No interpolation between stations.
Name frequency time series modeled as rise-peak-decay curves. Half-life extraction, contagion modeling, and trigger event detection via changepoint analysis.
Used in · The Given
M4Integration Impact Analysis
Career-adjusted WAR calculations across segregated and integrated eras. Fan chart trajectory modeling with 500 simulated career paths. Statistical record reconciliation.
Used in · The Color Line
M5Law–Outcome Regression
Panel regression modeling gun homicide rates against 134 state-level law provisions, with lagged effects, demographic controls, and difference-in-differences identification.
Used in · The Trigger
M6Immigration Wave Classifier
Time-series segmentation of arrival/deportation flows. Changepoint detection at policy boundaries. Quota system modeling with counterfactual projections.
Used in · The Crossing
M7Deinstitutionalization Transfer Model
Scissors-chart analysis tracking psychiatric bed closures against incarceration growth. State-level panel data with Granger causality tests for the transfer hypothesis.
Used in · The Untreated
M8Boundary Discontinuity Design
Regression discontinuity at HOLC grade boundaries. Properties within 500m of boundary lines compared with local polynomial regression. Causal estimates of redlining effects.
Used in · The Red Line
M9Consolidation Cascade Model
Ownership concentration tracking via FCC license transfer data. Market-by-market format diversity analysis pre/post Telecommunications Act of 1996.
Used in · Dead Air
All models are validated against published academic estimates where available. Confidence labels (HIGH / CANDIDATE / SPECULATIVE) are assigned per finding, not per model.
Data sources
Where the data comes from.
GHCN-Daily
NOAA / NCEI
12,847 stations
04 · The Atmosphere
NUFORC Archive
National UFO Reporting Center
111,961 reports
01 · The Sighting
SSA Baby Names
Social Security Administration
104,819 names
02 · The Given
Negro Leagues DB
Seamheads / Baseball Reference
2,300+ careers
03 · The Color Line
FBI UCR / SHR
FBI / BJS / CDC
1900–2024
05 · The Trigger
DHS Yearbooks
DHS / National Archives
1820–2024
06 · The Crossing
NIMH / SAMHSA / BJS
Federal mental health data
1900–2024
07 · The Untreated
HOLC Redlining Maps
Mapping Inequality / U of Richmond
239 cities · 1935–1940
09 · The Red Line
FCC Broadcast License DB
Federal Communications Commission
1934–2024
10 · Dead Air
License
All source code under MIT. Datasets under CC0 1.0 (public domain).
Raw data
Cleaned datasets in the GitHub repository under data/.