Every data point in every story is sourced from one of the datasets below. I analyse publicly available data. All raw data is available for download.
| Dataset | Organization | n | Method |
|---|---|---|---|
| Total Income by Age Group, Canada | Statistics Canada | Table 11-10-0239-01, average and median total income by age group and gender, 2024 constant dollars, 2015-2024. | |
| r-g Series and US Labour Share, 1950-2024 | FRED · Piketty & Zucman (2014) · World Inequality Lab | 74 | Real GDP growth (FRED, 5-year centred moving average) against net return on capital calibrated from Piketty-Zucman and the WIL 2025 update, 2-year lag. Correlated against BLS labour-share series. |
| Bachelors Without Bachelor's (NBER WP 35179) | Chambers, Goldman & Winkelmann, NBER | Marriage rates by birth cohort (1930-1980) from the paper's own figures, CPS-derived. Cross-checked against BLS CPS Table A-4 earnings by education. | |
| Team Shooting by Zone, 2001-2024 | Basketball-Reference | 716 | 716 team-seasons, shot distribution and FG% by zone. Teams ranked into within-season tertiles by analytics score. Season-level (N=24) and team-season regressions, clustered SEs by season. |
| PCE Inflation, Fed Funds, and Distributional Accounts | FRED · Federal Reserve Distributional Financial Accounts · PGIM | 52 | PCEPI, PCEPILFE, FEDFUNDS, DFII10 from FRED; wealth distribution from the Fed's Distributional Financial Accounts, Q4 2024; claims checked against PGIM's 2026 Mid-Year Global Market Outlook. |
| Predicted Incrementality by Experimentation (PIE) | Meta Research & Northwestern University | 2,226 | Gordon, Moakler & Zettelmeyer, March 2026. Last-click vs. PIE prediction accuracy against RCT ground truth across 2,226 real Meta ad experiments. |
| Player-Season xG and Non-Penalty Goals | Understat | 5,498 | 5,498 consecutive player-season pairs, six European leagues, 2014/15-2023/24. Min. 900 minutes and 0.05 xG/90 in both seasons of each pair. |
| Marriage Rates, Same-Sex Share, and Friend-Time | CDC/NCHS · Census ACS · Williams Institute · Pew · BLS ATUS | CDC/NCHS provisional marriage rates 2000-2023; Census ACS Table B11009 and Williams Institute for same-sex married share; Pew smartphone-ownership fact sheet; ATUS respondent file (TRTFRIEND, ages 20-34) 2003-2023, tested against a recent NBER paper on iPhone access and teen births. | |
| World Cup 2026 Pre-Kickoff Prices and Outcomes | Polymarket (Gamma & CLOB APIs) | 73 | Pre-kickoff implied probabilities, resolved outcomes, volume, and unique trading wallets for 73 matches, group stage through the semifinals. CLOB price history limited to a ~30-day rolling window. |
| Uplift Prediction Dataset | Criteo AI Lab | 13,979,592 | Diemert et al. (2018). ~14M users from real randomized ad-incrementality tests, treatment/control split, 12 anonymized features. Two-model uplift approach vs. plain response model, scored by Qini curve. |
| Legacy Admissions and Resume Audits | NBER · Economics of Education Review · American Sociological Review · ACLU | Harvard admit rates via Arcidiacono, Kinsler & Ransom (NBER WP 26316), from data unsealed in SFFA v. Harvard. Hurwitz (2011) legacy-admission odds across 30 selective colleges. Rivera (2012) culture-fit interviews and Rivera & Tilcsik (2016) resume audit (316 law firm offices, 14 cities). ACLU legacy-admissions tracking (top-100 universities). |