U.S. labor cracks showing
Labor market indicators are softening: JOLTS job openings fell to about 6.88 million in February and previews warned March payrolls could slow with rising unemployment. The signals point to weakening hiring momentum that quant economists should account for in nowcasting models. (fxstreet.com; cepr.net)
The JOLTS release shows hires fell to 4.8 million in February, a drop of 498,000 from the prior month, while quits held at 3.0 million and the quits rate was 1.9 percent, with openings down sharply in accommodation and food services (-211,000). (bls.gov) The JOLTS hires rate slipped to 3.1 percent in February — the lowest hires rate since April 2020 — a signal that employer hiring intensity has materially cooled even as total separations remained near 5.0 million. (bls.gov) Private payroll processors ADP reported +62,000 private‑sector jobs in March, with education and health services contributing 58,000 and construction adding 30,000, while small firms (<50 employees) accounted for +85,000 of the gains. (cnbc.com) Weekly claims data show initial jobless claims at about 210,000 for the week ending March 21, 2026, while continuing claims fell to roughly 1.82 million — a mixed signal that layoffs have not surged even as hiring softens. (money.usnews.com) A practical nowcasting setup for one‑month‑ahead payrolls: combine JOLTS (available Dec 2000–Feb 2026 on FRED) with weekly initial claims (DOL ICSA) and monthly ADP series, estimate a state‑space/dynamic factor model and backtest by training on 2005–2019 and testing on 2020–2025 to evaluate stability across pandemic and post‑pandemic regimes. (fred.stlouisfed.org) Reproducible student project: implement a Kalman‑filter DFM in Python using the large‑DFM Colab notebooks and the R nowcasting package, produce a daily-updated posterior predictive distribution for nonfarm payrolls, and document out‑of‑sample RMSE versus a baseline AR(1) across 2010–2025. (colab.research.google.com)
