The question
The pandemic hit the US labour market faster than any previous recession. How many jobs disappeared, how high did unemployment climb, and how long did it take to get back to where things stood in February 2020?
Approach
An R script calls the BLS Public Data API and requests three monthly, seasonally adjusted series:
- Unemployment rate (
LNS14000000), from the household survey; - Employment–population ratio (
LNS12300000), from the household survey; - Total nonfarm payrolls (
CES0000000001), from the establishment survey.
The API caps the span of a single request, so the script splits the period into 10-year windows. It uses the higher-limit v2 API when a key is present and falls back to v1 without one. The key is read from an environment variable, never from the code. The tidy result is written to a CSV, so every figure can be checked without calling the API again.
Results
| Indicator | Feb 2020 | Apr 2020 | Change | Back to pre-COVID level |
|---|---|---|---|---|
| Unemployment rate | 3.5% | 14.8% | +11.3 pp | May 2022 |
| Employment–population ratio | 61.1% | 51.2% | −9.9 pp | Not fully (59.2% in Sep 2026) |
| Nonfarm payrolls | 152.3 M | 130.4 M | −21.9 M jobs | June 2022, after 28 months |
Conclusions
- The fastest collapse in the series' history: roughly one in seven US jobs disappeared within two months.
- Jobs came back faster than people: payrolls recovered by mid-2022, but the share of the population in work never returned to its 2019 peak.
- Data quality is part of the story: the household survey was not conducted in October 2025 because of the federal shutdown, and the pipeline keeps that gap rather than interpolating a value.